scientific article; zbMATH DE number 4052997
[For the entire collection see Zbl 0638.00034.] One of the main disadvantages of functional languages is the slow execution of functional programs. By choosing and designing the right reduction model (lambda-calculus, combinators, rewrite systems), the right machine architecture (sequential/parallel reduction machines) and by using clever compiling techniques one tries to solve this problem. One of the fastest sequential reduction machines is the G-machine. In this paper we will show how this concept can be extended to a parallel version. The parallelism in this machine will be defined by serial combinators [\textit{P. Hudak} and \textit{B. Goldberg}, Lect. Notes Comput. Sci. 201, 382-399 (1985; Zbl 0568.68018)]. Based on these concepts a simulator is built. This simulator is used to obtain some measurements of the performance and behaviour of the parallel G-machine under various circumstances. It turns out that the amount of parallelism reached is sufficient, while the overhead produced is acceptable. However, the machine architecture simulated is not a very good one.
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3789053)