Asynchronous Parallel Discontinuous Finite Element Method

We describe a new iterative, asynchronous, parallel algorithm for the solution of partial differential equations, based on discontinuous finite-element methods. We use the domain-decomposition methods to decompose a large problem into a number of smaller problems that can be computed in parallel. These methods facilitate coarse-grain parallelism, which is important for exploiting parallelism efficiently. Numerical experiments that were executed on the MOSIX cluster computing system, show the new algorithm to be robust and highly parallelizable, with an almost linear speedup with respect to the number of processors.