Research on parallel model for sparse matrix-vector iterative multiplication

The most effective algorithms of solving large sparse linear system are Block Wiedemann and Block Lanczos, sparse matrix-vector multiplication iterations is the main process of these algorithms, to achieve parallel computing of its process, we have established three different parallel algorithm models and analyzed the characteristic features of their computing and communication, and through the analysis and comparison of time-cost, we choose the optimal model. © 2013 IEEE