Toward the Concurrent Implementation of Computational Systems

This paper places concurrent rewriting in the perspective of supporting the development and use of computational systems. We briefly recall the usefulness of computational systems (i.e. a rewriting theory with strategies) as a logical framework. We then show that the right implementation model for concurrent rewriting allows us to give a meaning to the concurrent rewrite of overlapping redexes. The principles of the implementation of concurrent rewriting and of the associated garbage collection are sketched. We then present a specific transformation of conditional rewriting to unconditional rewriting, that permits us to preserve the potential concurrency of the system as much as possible. This allows us to extend the ReCo system to handle conditional specifications. We finally present experimental results of the system on various parallel architectures.