Abstract

It is a time consuming and very skilful task for researchers or developers in computational mechanics to modify a program, designed for a single processor, to one suitable for parallel computation. This is a serious bottleneck in parallel computation, even though a general-purpose parallel computational library, such as MPI, is applied to this modification. We have developed a parallel matrix solver platform, based on a domain decomposition method, for various numerical schemes such as the finite element method (FEM), the finite difference method and the finite volume method, to accelerate a smooth shift to the realm of parallel computation. Parallel software such as PETSc, Aztec, GEOFEM and ADVENTURE have already been developed, however these systems are more suitable for professionals in parallel computation and not valid for our purpose. In our platform, a user is merely required to call the platform at the stage of stiffness matrix calculation. GMRES and Bi-CGSTAB with several pre-conditioners are used as a basic matrix solver. The option of invoking a Lagrange-multiplier is also included. For partitioning, a fast graph generator for arbitrary elements and an interface with MeTis are provided. Our platform is valid for a variety of hardware, including a single processor based workstation, through the exchange of Makefilein. The effectiveness of our platform is evaluated with several examples in the area of finite element fluid dynamics in this paper.

Full Text
Paper version not known

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