Parallelization of a finite element Fortran code using OpenMP library

摘要

A finite element analysis code FEAP is parallelized using the OpenMP library, and the performance is evaluated. The essential concepts of the OpenMP for the implementation of the parallelization are described and then our parallelization procedures are presented. The skeleton of FEAP is analyzed and the subroutines that need to be parallelized are identified. Then, the general procedures to declare the parallel variables are explained and synchronization, which is necessary for parallelization, is described. Using the parallelized FEAP code, several examples are solved, and the accuracies and the numerical efficiencies are investigated.