The parallel iterative algorithm based on approximate factorization for block-tridiagonal linear systems

摘要

In this paper, the parallel iterative algorithm for solving block-tridiagonai linear systems on distributed-memory multi-computers is presented. The coefficient matrix of the block-tridiagonal linear systems is approximately decomposed in order to perform the parallel computation, and an iterative scheme is formed. The communication needs only thrice between the adjacent processors all through the computing process at each iteration. Theoretically, a sufficient condition of convergence of the algorithm is given.Finally, the results on HP rx2600 cluster indicate that the algorithm is feasible with the preferable convergen -ce and efficiency.