基于多核集群的RTI并行优化技术

Parallelization Techniques for Multi-Core-Cluster-based RTI

随着基于高层体系结构HLA(High Level Architecture)仿真应用的不断深入,系统仿真规模不断扩大,仿真模型越来越复杂,仿真节点之间的交互更加频繁,使得仿真系统对通信资源与计算资源的需求不断提高。多核集群能够提供数据并行处理和高效通信的性能潜力,因此通过在多核集群上进行仿真加速正成为提高HLA仿真运行效率的重要途径。然而传统的仿真运行支撑平台RTI(Run Time Infrastructure)主要面向分布式网络环境,一般不支持并行通信和消息并行处理,从而难以发挥多核集群的高效通信和并行计算的性能优势。因此,为了提高多核集群下大规模HLA仿真的运行效率,提出了基于多通道的并行通信结构和基于消息分类的多线程处理技术,并已经应用于正在开发的并行shmRTI中,与国际同类商业软件pRTI1516进行的对比测试结果表明,性能可提升30%以上。

With the deepening of HLA （High Level Architecture）-based simulation applications, the scale of System Simulation continues to expand and the simulation has become increasingly complicated. As a result, more frequent interaction occurs between the simulation nodes, and the demand of simulation system for communication and computing resources increases. Multi-Core-Cluster has more potential for performance of parallel processing and efficient communication, so accelerating simulation in the Multi- Core-Cluster is becoming an important way to improve the efficiency of HLA simulation run. However, the traditional simulation run support platform of RTI （Run Time Infrastructure） is mainly for the distributed network environment, and generally does not support parallel communication and message parallel processing, making it difficult to take advantage of the Multi-Core-Cluster on efficient communication and parallel computing. Therefore, in order to improve the operating efficiency of large-scale HLA simulation in Multi-Core-Cluster, the multi-channel-based parallel communication structure and the message-classification-based multithreading processing technology was proposed. Both have been used in the parallel shmRTI. Compared with similar commercial software pRTI1516, the test results show that performance of shmRTI is more than 30 percent better than that of pRTI 1516.