有限元的OpenMp并行计算技术

传统有限元串行计算技术耗时长、效率低，已远远不能满足工程实践需要。通过分析串行计算的耗时分布、程序可并行性及数据相关性，在保持串行程序不变的基础上，采用基于多线程的OpenMP技术实现单元刚度矩阵形成总体刚度矩阵的程序并行化。针对有限元计算机翼实例进行了不同线程数目下并行加速比的研究，结果表明，OpenMP技术可有效提高有限元计算效率，且存在一个最佳的线程数，使得并行加速比达到最大，此外加速比还与问题的计算量等因素有关。

Parallel Calculation Technology for Finite Element Based on OpenMP

Traditional serial finite element calculation is time-consuming and low-efficient,which hasn＇ been able to satisfy requirements of engineering practice. Based on multi- thread technology, OpenMP can make the assembly of global stiffness matrix from element stiffness matrices parallel without modifying the original serial program by analysis of consumed time distribution of serial calculation,program＇s parallel- ization and the data correlation. A conclusion can be drawn from the study of speedups under different number of threads in the finite element analysis of a wing structure that OpenMP can improve the efficien- cy of finite element calculation significantly, and an optimal number of threads exists, corresponding to a maximum parallel speedup. Additionally, the parallel speedup is dependent on the amount of calculation of the problem to be dealt with and so on of the problem to be dealt with and so on.