摘要
构建了一种适用于多核集群的混合并行编程模型.该模型融合了共享内存的面向任务的TBB编程和基于消息传递的MPI编程两种模式.结合两者的优势,实现进程到处理节点和进程内线程到处理器核的两级并行.相对于单一编程方式下的程序性能,采用这种混合并行编程模型的算法不但可以减少程序执行时间,获得更好的加速比和执行效率,而且明显地提高了集群性能.
To take full advantage of the structural characteristics of multi-core clusters and enhance the use of CPU each core efficiently,a multi-core cluster hybrid parallel programming model is constructed.The model combines the shared memory and the TBB task-oriented programming and MPI message-passing programming mode.Using their advantages,achieve the hierarchical parallel programming of processes to nodes and threads to the processor core.Compare with the algorithm of the single parallel programming model,this programming mode not only can reduce the program execution time,get a better speedup and execution efficiency,but also improve the cluster performance greatly.
出处
《微电子学与计算机》
CSCD
北大核心
2011年第2期91-93,97,共4页
Microelectronics & Computer
基金
国家自然科学基金项目(60603047)
辽宁省教育厅高等学校科研基金项目(05L209)
