A benchmark for artificial intelligence applications on parallel computers-BEAP

Production or rule based systems hold an important place within the field of artificial intelligence (AI). In present times production systems are facing a bottleneck regarding speeding up of their processing. The use of parallel computers has been advocated as one of the most appropriate technologies for this speedup. In this direction, attempts have been made and more will follow. These attempts need to be measured. Measurement through benchmarks happens to be the most accurate way of making quantitative product analysis. While research is being carried out to find complete range of benchmarks for all kinds of application, somehow, work on benchmarks for AI applications have been left behind. This work fills that gap. The paper has three purposes: (1) to define what is a benchmark and what constitutes a good benchmark; (2) to describe the characteristics of AI applications on parallel computers and to bring out how parallelism can be exploited; and (3) to outline a benchmark for AI applications on parallel computers.