Communication results for parallel sparse Cholesky factorization on a hypercube (Q1263230)

In recent years, parallel algorithms have been developed for solving sparse positive definite systems. In this paper, the authors have considered the problem of reducing data traffic between processor nodes during the execution of a parallel algorithm. A task assignment strategy based on the structure of an elimination tree is considered. The paper also contains some numerical experiments, which show that the new allocation scheme does reduce communication traffic and leads to improvements in the performance.