The Berland Army is preparing for a large military parade. It is already decided that the soldiers participating in it will be divided into k rows, and all rows will contain the same number of soldiers.
Of course, not every arrangement of soldiers into k rows is suitable. Heights of all soldiers in the same row should not differ by more than 1. The height of each soldier is an integer between 1 and n.
For each possible height, you know the number of soldiers having this height. To conduct a parade, you have to choose the soldiers participating in it, and then arrange all of the chosen soldiers into k rows so that both of the following conditions are met:
each row has the same number of soldiers,
no row contains a pair of soldiers such that their heights differ by 2 or more.
Calculate the maximum number of soldiers who can participate in the parade.
Input
The first line contains one integer t (1≤t≤10000) — the number of test cases. Then the test cases follow.
Each test case begins with a line containing two integers n and k (1≤n≤30000, 1≤k≤1012) — the number of different heights of soldiers and the number of rows of soldiers in the parade, respectively.
The second (and final) line of each test case contains n integers c1, c2, …, cn (0≤ci≤1012), where ci is the number of soldiers having height i in the Berland Army.
It is guaranteed that the sum of n over all test cases does not exceed 30000.
Output
For each test case, print one integer — the maximum number of soldiers that can participate in the parade.
Example
Input
5
3 4
7 1 13
1 1
100
1 3
100
2 1
1000000000000 1000000000000
4 1
10 2 11 1
Output
16
100
99
2000000000000
13
Note
Explanations for the example test cases:
the heights of soldiers in the rows can be: [3,3,3,3], [1,2,1,1], [1,1,1,1], [3,3,3,3] (each list represents a row);
all soldiers can march in the same row;
33 soldiers with height 1 in each of 3 rows;
all soldiers can march in the same row;
all soldiers with height 2 and 3 can march in the same row.
思路:这是一个单调函数,具有单调性,因此我们可以二分每一行的人数,判断是否符合条件。在判断的时候,我们贪心的去想,前面的尽量用,以免后面不够的情况。在判断的时候,我是用的dfs,有一点要注意,如果当前值可以满足当前行的人数,那么我们就直接算出这一标号的人可以满足多少行,否则会RE的,因为K太大了。
代码如下:
#include<bits/stdc++.h>
#define ll long long
using namespace std;
const int maxx=3e4+100;
ll a[maxx];
int n;
ll k,sum;
inline void dfs(ll &row,int &flag,int pos,ll x)
{
if(flag==0) return ;
if(row>=k+1) return ;
if(pos==n+1)
{
flag=0;
return ;
}
if(a[pos]>x)
{
ll t=a[pos]/x;
a[pos]%=x;
dfs(row+=t,flag,pos,x);//这里要注意
a[pos]=t*x+a[pos];//这个操作类似于还原数组,这样就不用每一次都遍历数组还原了。
}
else if(a[pos]==x) dfs(row+=1,flag,pos+1,x);
else
{
if(a[pos]+a[pos+1]<x)
{
dfs(row,flag,pos+1,x);
return ;
}
a[pos+1]-=(x-a[pos]);
dfs(row+=1,flag,pos+1,x);
a[pos+1]+=(x-a[pos]);
}
}
inline int check(ll x)
{
if(sum<x*k) return 0;
int flag=1;ll row=1;
dfs(row,flag,1,x);
return row>=k+1&&flag;
}
int main()
{
int t;
scanf("%d",&t);
while(t--)
{
sum=0;
scanf("%d%lld",&n,&k);
for(int i=1;i<=n;i++) scanf("%lld",&a[i]),sum+=a[i];
a[n+1]=0;
ll l=1,r=1e16,mid,ans=0;
while(l<=r)
{
mid=l+r>>1;
if(check(mid)) l=mid+1,ans=mid;
else r=mid-1;
}
printf("%lld\n",ans*k);
}
return 0;
}
努力加油a啊,(o)/~