A PARALLEL FINITE-ELEMENT METHOD WITH A SUPERCOMPUTER NETWORK

Computer simulations are about to replace experiments in various field s, and the scale of the models to be simulated tend to be extremely la rge. To perform large-scale finite element analyses, the authors propo se the parallel use of multiple supercomputers connected to one anothe r through a high-speed network. In other words, a supercomputer networ k is regarded as a parallel computer. As a parallel numerical algorith m for the finite element analysis, we adopt the domain decomposition m ethod (DDM) combined with an iterative solver, i.e. the conjugate grad ient (CG) method, where a whole analysis domain is fictitiously divide d into a number of subdomains without overlapping. Finite element anal yses of the subdomains are performed under the constraint of both disp lacement continuity and force equivalence among subdomains. Such a con straint can be satisfied through iterative calculations such as the CG method. The present DDM-based parallel finite element algorithm is co mbined with the server-client model for data and processor management to have the workload balanced dynamically between the processors, and is implemented first on an engineering workstation (EWS) network and t hen on a supercomputer network. The accuracy and parallel performance of the present system are tested using the network composed of various EWSs. Finally, it is demonstrated that the present system implemented on the supercomputer network can solve the three-dimensional elastici ty problem of over one million degrees of freedom at an extremely high average speed of 1.74 GFLOPS.