Abstract: The finite element method is used extensively in the solution of complex engineerin;: problems. In this paper, a parallel finite element method and its implementation on a two dimensional hyper -ring supercomputer system are described- The proposed strategy does not require the formation of an integrated global system equations and each processor stores only the information relevant to its element (or substructure) and generates the element stiffness matrix and assembles the substructure stiffness matrix in parallel, A parallel LU decomposition scheme is employed to solve the system equations. Parallel architectures of the systems have significant influence on the parallel algorithm. In this paper, it is demonstrated that for finite element computation, using a two-dimensional hyper-ring system is preferred to a hypercube system. Moreover; we have studied the problems of the task partitioning and the task assignment of the parallel finite element method. This method can significantly reduce the execution time for a large scale finite element problem.

Key words: parallel tintte element method paprullei computer, performance evaluation, task partrtioning, task assignment