Large numerical linear algebra in 1994: the continuing influence of parallel computing

The note covers two aspects of the state of the art of large numerical linear algebra problems. Firstly, we look at the current records for sparse and dense linear systems and eigenvalue problems on a variety of machines. The second subject matter is perhaps more of a question than an answer; we explore why network topologies of a parallel machine are hardly ever used in ways that perhaps a graph theorist might envision, especially given that linear algebra, particularly dense linear algebra, consists of many very regular predictable operations