The impact of cache coherence protocols on parallel logic programming systems

Abstract

In this paper we use execution-driven simulation of a scalable multiprocessor to evaluate the performance of the Andorra-I parallel logic programming system under invalidate and update-based protocols. We use two versions of Andorra-I. One of them was originally designed for bus-based multiprocessors, while the other is optimised for scalable architectures. We study a well-known invalidate protocol and two different update-based protocols. Our results show that for our sample logic programs the update-based protocols outperform their invalidate-based counterpart for the original version of Andorra-I. In contrast, the optimised version of Andorra-I benefits the most from the invalidate-based protocol, but a hybrid update-based protocol performs as well as the invalidate protocol in most cases. We conclude that parallel logic programming systems can consistently benefit from hybrid update-based protocols.