Reflections on parallel functional languages

Abstract

Are parallel functional languages feasible? The large majority of the current projects investigating this question are based on MIMD machines and the current set of implementation methods for functional languages which is graph rewriting and combinators. We regret that we have to come to a negative conclusion: it is highly unlikely that these projects will turn parallel processing by means of functional languages into an attractive alternative to sequential processing, not even for special purposes. Although we express a clear conclusion, we have included some responses from the projects mentioned in this paper to allow the reader to make up his own opinion.