常见的实现方法是使用两个数组，

t[n-1]和e[n],数组e存放所有的参加排序的元素，在完全二叉树中是叶子结点，t[]存放

的是比赛的阶段性胜者的编号，再进行n趟比赛排序结束，时间复杂度是O(n*log2(n))，

其实我觉得也可以只通过一个数组A[2*n-1]也可以实现，只是标号的换算有点复杂而已.

-Common implementation approach is to use two array, t [n-1] and e [n], the array storage of all e sort of take part in the elements, are in complete binary tree in leaf nodes, t [] are stored stage game winner serial number and then sort the end of n times a game, time complexity is O (n* log2 (n)), in fact I think it is only through an array A [2* n-1] can also be implementation, but labeling it a bit complicated conversion.