Abstract: To well take the advantage of performance of multi-core processor, aiming at hierarchical Network on Chip(NoC) architecture, this paper presents the MPI/OpenMP hybrid parallel programming model. The model uses MPI-based task-level parallel model to achieve efficient communication between clusters. It uses OpenMP method to implement communication, synchronization and data exchange between four cores inside clusters. Experimental results show that the model can increase the speedup by 20% to 50% compared to the single parallel programming model.

Key words: Network on Chip(NoC), Message Passing Interface(MPI), OpenMP model, hierarchical, hybrid parallel programming

摘要： 为更好发挥多核处理器的硬件性能，针对层次化的片上网络架构，提出MPI/OpenMP混合并行编程模型。运用基于MPI的任务级并行模型实现片内簇间的高效通信，采用OpenMP模型实现簇内四核的通信、同步和数据交换。实验结果表明，与单一并行编程模型相比，混合并行编程模型加速比提高了20%~50%。

关键词: 片上网络, 消息传递接口, OpenMP模型, 层次化, 混合并行编程

CLC Number: 

• N945.12