摘要
针对高速碰撞仿真系统中计算量较大的有限元计算部分,建立了并行有限元模型和区域划分原则,利用逐元法(Element by Element,EBE)分析有限元结构和单个有限元的物理量,利用区域划分方法对有限元网格进行划分,并采用中心差分法求解拉格朗日运动方程,用以模拟圆柱形物体高速碰撞刚性墙的过程。分析了多核CPU架构的基本通信原理,提出了1种动态负载均衡策略,能够优化多核架构上高速碰撞过程中有限元计算的运行效率,选用MPICH2和FORTRAN编写了仿真计算程序。经过多次实验,该优化方法将有限元计算程序运行效率平均提高了8.7%以上。
Aimed at the finite element calculation part with high computational complexity in high tem, we establish a parallel finite element model and regional partition principle, analyze the structure and nite element by Element by Element(EBE) method’ divide finite element mesh by regional partition principle, solve Lagrangianmotion equation by central difference method, and simulate the process of cylinder impact rigid walls. By Analyzing the basicprinciple of inter-core communication undermulti-core architecture, the dynamic load balancing strategy is given to optimize theoperation efficiency of finite element program under multi-core architecture, finite element program is programed by MPICH2standard library and FORTRAN language. Multipleexperiments show that our optimization method average increased thepro-gram operating efficiency more than 8. 7%.
作者
崔浩龙
魏晓辉
徐海啸
CUI Haolong;WEI Xiaohui;XU Haixiao(College of Computer Science and Technology, Jilin University, Changchun 130012, China;High Performance Computing Center,College of Computer Science and Technology,Jilin University, Changchun 130012, China)
出处
《中国科技论文》
北大核心
2017年第20期2295-2300,共6页
China Sciencepaper
基金
国家重点研发计划专项(016YFB0201503
2016YFB0701101)
国家自然科学基金资助项目(61602205
51627805
61170004)
高等学校博士学科点专项科研基金资助项目(20130061110052)
关键词
计算机应用
并行程序优化
信息传递接口
有限元技术
computer application
parallel program optimization
message-passing-interface(MPI)
finite element technology