SCALABLE ALGORITHMS FOR FAULTY R-MESHES

摘要

Recently, many parallel computing models using dynamically reconfigurable electrical buses have been proposed in the literature. All algorithms developed for these models assume that a healthy system is available. Unfortunately, these algorithms are useless if we assume that one processor fails in the system. To overcome this problem, this paper proposes a new approach to develop scalable fault-tolerant algorithms for the Reconfigurable Mesh. The proposed approach uses standard techniques and enables the execution of any algorithm on an NxN R-Mesh with at most N faulty processors. This is accomplished by utilizing scalable algorithms incurring a low overhead in time.