Some aspects of parallel implementation of the finite-element method on message passing architectures (Q908670)

In the finite element method it is usual to handle with large matrices. Therefore it is convenient to ask for the employment of a new computer generation, based on the emergence of large-scale parallel architectures. The authors present an analysis of the solution techniques in a model case using three versions of the finite-element method: the standard h- version, the p-version and the h-p-version as discretizations inside super-elements including direct local elimination and global elimination by both direct methods and iterative methods. In their conclusion they give hints for the application of sparse direct solvers which seem to be useful especially to systems arising from both the h- and h-p-method.