Performance of Lisp systems

Abstract

This paper describes the issues involved in evaluating the performance of Lisp systems. We explore the various levels at which quantitative statements can be made about the performance of a Lisp system, giving examples from existing implementations wherever possible. Our thesis is that benchmarking is most effective when performed in conjunction with an analysis of the underlying Lisp implementation and computer architecture. We examine some simple benchmarks which have been used to measure Lisp systems, and examine some of the complexities of evaluating the resulting timings.