Search results

Computational algorithms for finite element dynamic analysis of large‐scale structural problems that exploit both concurrent and parallel features of multiple instruction multiple data streams computers are presented. A new computer program architecture is used in which large finite element domains are automatically divided into subdomains. The number of subdomains generated is equal to the number of available processors. The spatial solution is obtained using a basis of orthogonal vectors. The temporal solution is computed exactly. Discussion is focused on the concurrent generation of global Ritz vectors. Examples run on a hypercube multiprocessor confirm the potential of the proposed scheme.

The various forms of parallel numerical algorithms that speed up finite element computations are as numerous as the number of researchers working on the problem. In this paper, we review some of these parallel computational strategies and assess their adequacy for a given architecture and a given problem. We also report on the performance of both extreme parallel hardware technologies on real‐life structural problems.