并行化思路实现dijkstra算法。假设有p个处理器，N个顶点。给每个处理器分配N/p个顶点，求出局部的最小值，复杂度为O (N/p)。然后后一半的处理器将自己的最小值发送给第前p/2个处理器。前一半处理器接收到传来的值后，与局部的最小值比较，作为新值。继续循环，直到剩下一个处理器为止。-Parallelized dijkstra algorithm. Suppose there is p a processor, N vertex. Give each processor distribution N/p a vertex, find out the local minimum value, complexity is O (N/p). And then later half of the processors will sent its own minimum value to the first former p/2 processors. the first half receive the value of the processor and compare with the local minimum value,as a new value. Continue the circulation, until only one processor is left.