Scalability analysis of parallel GMRES implementations

Applications involving large sparse nonsymmetric linear systems encourage parallel implementations of robust iterative solution methods, such as GMRES(k). Two parallel versions of GMRES(k) based on different data distributions and using Householder reflections in the orthogonalization phase, and variations of these which adapt the restart value k, are analyzed with respect to scalability (their ability to maintain fixed efficiency with an increase in problem size and number of processors). A theoretical algorithm-machine model for scalability is derived and validated by experiments on three parallel computers, each with different machine characteristics.