The fastest segment tree is the self-adjusting one according to the queries.
The root of self-adjusting segment tree is , and for each vertex , we can choose arbitrary and then split it into two vertices and . The following figure is an illustration of such a segment tree.

When we locate an interval in the segment tree, we need to visit some vertices. For example, the vertices in purple rectangles are visited when locating the interval .
Formally, when we locate the interval , let be the set with minimum number of vertices such that their intervals are disjoint and the union of there intervals is , all the vertices in S or the ancestors of some vertex in will be visited.
Now you are given queries  and you need to find a segment tree that minimizes the sum of visited vertices for all queries.

The first line contains two integers .
Each of the following  lines contains two integers .

Output one integer - the minimum sum of visited vertices for all queries.

5 5
4 5
2 4
2 5
1 5
1 2

16