On the Parallelization of Domain Decomposition Methods for 3-D Boundary Value Problems

An investigation of parallel domain decomposition methods for interactive solution of 3D boundary value problem is presented. The various variants of algorithms are considered: the different values of subdomain overlapping, the different numbers of subdomains and processors, accelerated and non-accelerated two-level iterations. The dependence of speedup on the computational parameters is discussed on the base of numerical experiment at the multiprocessor computer Fujitsu-Siemense RM600.