Abstract

Most of the calculation time for large scale numerical simulation which uses rigid-plastic finite element method is occupied by the solver part of simultaneous linear equation called stiffness equation.We incorporated the parallelized conjugated gradient method onto our cluster of workstations to shorten the computation time of them.It is important for high speed parallel computation in our cluster system of many workstations to reduce the data communication time through TPDDI network.In this paper, we propose a new method which enables to reduce data communication time by using the skewed tree multiple data transfer through concurrent or parallel paths from plural slave processors to that of others like systolic pulsations. Then, we achieve about eightfold speedup of execution time by twenty-four slave processors.And, we infer that the method shown in this paper is also effective to elastic finite element method.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call