This paper on "The Efficient Use of Parallel Computing with PERMAS" was presented at the NAFEMS World Congress on Design, Simulation & Optimisation: Reliability & Applicability of Computational Methods - 9-11 April 1997, Stuttgart, Germany.

Abstract

The general purpose Finite Element system PERMAS has been extended to support shared and distributed parallel computer architectures as well as workstation clusters. The methods used during the parallelization of this large application software package are of high generality and have the capability to parallelize all mathematical operations in a FE analysis - not only the solver.
Utilizing the existing hyper-matrix data structures for large, sparsely populated matrices, a programming tool called PTM was introduced that totally hides parallelization from higher order algorithms. With PTM an operation graph of sub-matrix operations is asynchronously build and executed. A clustering algorithm distributes the work, performing a dynamic load balancing and exploiting data locality.
The generality of the approach is presented also by benchmark examples dealing with different kinds of analysis, such as linear static, dynamic and contact analyses as well as electromagnetic wave propagation.