codeforces 61E

 Enemy is weak

time limit per test

5 seconds

memory limit per test

256 megabytes

input

standard input

output

standard output

The Romans have attacked again. This time they are much more than the Persians but Shapur is ready to defeat them. He says: "A lion is never afraid of a hundred sheep".

Nevertheless Shapur has to find weaknesses in the Roman army to defeat them. So he gives the army a weakness number.

In Shapur's opinion the weakness of an army is equal to the number of triplets i, j, k such that i < j < k and ai > aj > ak where ax is the power of man standing at position x. The Roman army has one special trait — powers of all the people in it are distinct.

Help Shapur find out how weak the Romans are.

Input

The first line of input contains a single number n (3 ≤ n ≤ 106) — the number of men in Roman army. Next line contains n different positive integers ai (1 ≤ i ≤ n, 1 ≤ ai ≤ 109) — powers of men in the Roman army.

Output

A single integer number, the weakness of the Roman army.

Please, do not use %lld specificator to read or write 64-bit integers in C++. It is preffered to use cout (also you may use %I64d).

Sample test(s)

input

3
3 2 1

output

1

input

3
2 3 1

output

0

input

4
10 8 3 1

output

4

input

4
1 5 4 3

output

1

#include <iostream>
#include <stdio.h>
#include <algorithm>
#include <cstring>
#define MAX 1000100


using namespace std;


struct node
{
    long long val;
    int pos;
}a[MAX];


long long ans[MAX],ans2[MAX];
long long  c[MAX];


bool cmp1(node a,node b)
{
    return a.val<b.val;
}


bool cmp2(node a,node b)
{
    return a.pos<b.pos;
}


long long lowbit(int x)
{
    return x&(-x);
}


void add(int po,long long val)
{
    while (po<=MAX)
    {
        c[po]+=val;
        po+=lowbit(po);
    }
}


long long getsum(int po)
{
    long long sum=0;
    while (po>0)
    {
        sum+=c[po];
        po-=lowbit(po);
    }
    return sum;
}


int main()
{
    int n;
    freopen("in.txt","r",stdin);
    while (cin>>n)
    {
        if (n==0) break;
        int i;
        for (i=1;i<=n;i++)
         {
            scanf("%lld",&a[i].val);
            a[i].pos=i;
         }


        sort(a+1,a+n+1,cmp1);          //离散化
        for (i=1;i<=n;i++) a[i].val=i;
        sort(a+1,a+n+1,cmp2);


        memset(c,0,sizeof(c));       //求二元的
        memset(ans,0,sizeof(ans));
        for (i=1;i<=n;i++)
        {
            add(a[i].val,1);
            ans[i]=i-getsum(a[i].val);
         }


        memset(c,0,sizeof(c));        //求三元的
        memset(ans2,0,sizeof(ans2));
        long long sum=0;
        for (i=1;i<=n;i++)
        {
            add(a[i].val,ans[i]);
            sum+=ans[i];             //用所有项的总和减去比a[i]小的总和，即的比a[i]大的总和
            ans2[i]=sum-getsum(a[i].val);
        }


       long long  fians=0;
       for (i=1;i<=n;i++) fians+=ans2[i];
       cout<<fians<<endl;


    }
    return 0;
}
/*
  The first line of input contains a single number n (3 ≤ n ≤ 106)
   Next line contains n different positive integers ai (1 ≤ i ≤ n, 1 ≤ ai ≤ 109)
 求三元逆序对
 树状数组：
 设ans(i)为1到i-1中比a[i]大的元素个数,则显然逆序对数为(∑ans(i),i∈[2,n]).
 我们不妨再设一个ans2(i)=∑ans(j)(j∈[2,i-1],A[j]>A[i]),这样,ans2(i)就成为以i结尾的三元逆序组的个数,
 则最终答案就是(∑ans2(i),i∈[3,n]).
 同理,我们可以迭代计算cardbigger3,cardbigger4,cardbigger5,...,从而计算出K元逆序组.
 注意,所有的操作是串联的,也就是说计算K元逆序组只需要把以上算法重复K-1次,所以计算复杂度为O(knlogn),
 在k很小的时候则可看作是O(nlogn),算是很高效了.
 本体数据量很大，用long long,离散化
 */