Resolving Numerical Anomalies in Scientific Computation

The increasing numbers of large scientific computer systems in use, together with the rapidly increasing scale of computations being attempted on such systems, means that numerical anomalies pose a growing threat to the soundness of these computations. This note presents a number of examples of common numerical anomalies that can arise, illustrated by some actual computer programs available on a website, together with a discussion of how these problems could be mitigated with improved software tools. ∗Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA, [email protected]. Supported in part by the Director, Office of Computational and Technology Research, Division of Mathematical, Information and Computational Sciences, U.S. Department of Energy, under contract number DE-AC02-05CH11231.