针对目前图形处理器(GPU) 上的动态数据处理问题，在分析现有并行单源最短路径(SSSP) 算法的基础上，对 GPU上的Moore SSSP算法进行并行化设计与实现。搜索时，综合应用层次化任务分配、层次化工作队列、层次化 Kernel 调用等策略。在不同类型图数据上进行实验测试，实验结果表明，该算法能有效减少空线程开销、访存开销以及同步时间。-For dynamic data processing on the graphics processor (GPU), the analysis of existing parallel single-source shortest path (SSSP) algorithm based on the Moore SSSP algorithm on GPU Parallel design and realization. Search application level task allocation, hierarchical work queue, hierarchical Kernel call strategy. Experimental tests on the different types of map data, experimental results show that the algorithm can effectively reduce the empty thread overhead and memory access overhead and synchronization time.