国产欧美精品一区二区,中文字幕专区在线亚洲,国产精品美女网站在线观看,艾秋果冻传媒2021精品,在线免费一区二区,久久久久久青草大香综合精品,日韩美aaa特级毛片,欧美成人精品午夜免费影视

Parallel problem and shortest path problem has become a hot research topic, traditional shortest path algorithm cannot meet the demand of the explosive growth of the data processing, especially when the network size is large, the computation time and storage space required is greatly increased.The emergence of MapReduce model, brings a new solution to solve the shortest path. GPU has powerful parallel computing capability and storage bandwidth, and CPU has obvious advantages.By studying MapReduce model and GPU implementation process analysis, pointed out the shortest path parallel method based on MapReduce model alone existing problems, and reduce the performance of the system.The innovation of this paper is combine MapReduce and GPU to form double parallel model, parallel preprocessing data, the data transfer and synchronization overhead for the shortest patht,increase data dynamic processor. Compared with the average speedup of performance evaluation index of parallel algorithm, the results show that the computation of the shortest path in double parallel environment improves the speedup.