Application of domain decomposition methods for numerical modeling of wave phenomena on serial and parallel computer architectures

Domain decomposition has significant potential for analyzing large problems. By partitioning a problem into logical blocks, repetition can be avoided and efficiency can be increased. The most applicable solution technique can be used for each block without regard to the behavior in other blocks. Large and complex structures can be conveniently analyzed piecewise. Domain decomposition is well suited for massively parallel processing, although special consideration is necessary to maximize the efficiency. The technique demonstrates that elementary tasks such as boundary value problems can be solved in new ways that take full advantage of parallelism. The paper explains the basics of the technique, and gives a few examples of its application