摘要
为了提高机载设备振动环境实测数据处理效率,提出了一种并行化数据处理和振动环境谱编制方法。在SMP(symmetrical multi-processing)集群系统的多级体系结构下,以Welch(改进周期图法)算法为基础,进行了振动数据处理模型的并行化分析,对于并行化过程中存在的并行I/O、负载平衡等关键问题进行了讨论,提出了相应的解决方案。最终选择基于MPI/Open MP混合方式实现了算法。在搭建的集群环境下测试表明,12核参与运算的条件下,最高加速比可以达到7.4,有效提高了运算效率。
The parallel method was presented for enhancing data processing efficiency of vibration environment for airborne equipment. Under the memory hierarchy of the SMP cluster and base on parallel Welch method, Paral- lel analysis of vibration data processing model was completed. In addition, Some critical factors affecting the parallel performance such as parallel I/O, load balance were discussed and the corresponding solutions were proposed. Implement the algorithm by the way of MPI/OpenMP hybrid parallel ultimately. The Test shows that a highest speedup of 7.4 in 12 cores can be archived. Therefore algorithms can improve operation efficiency greatly.
出处
《科学技术与工程》
北大核心
2015年第20期72-77,115,共7页
Science Technology and Engineering
基金
国家自然科学基金项目(50135010)资助
关键词
并行算法
振动环境谱
功率谱估计
数据归纳
MPI
OPENMP
parallel algorithm vibration environment spectrum power spectrum estimation data inductive method MPI OpenMP