Flexible scheduling of or-parallelism in Aurora: The Bristol scheduler

Abstract

Aurora is a prototype or-parallel implementation of the full Prolog language for shared memory multiprocessors, based on the SRI model of execution. It consists of a Prolog engine based on SICStus Prolog and several alternative schedulers. The task of the schedulers is to share the work available in the Prolog search tree

This paper describes the Bristol scheduler. Its distinguishing feature is that work is shared at the bottom of partially explored branches (“dispatching on bottom-most”). This can be contrasted with the earlier schedulers, which use a “dispatching on topmost” strategy. We argue that dispatching on bottom-most can lead to good performance, by reducing the overheads of scheduling.

Our approach has been to find the simplest scheduler design which could achieve performance competitive with earlier more complex schedulers. This design gives us a flexibility in deciding strategies for sharing work and allows us to examine ways of improving the performance on both non-speculative and speculative work. We note that in speculative regions the priority of some work is higher than others. We have investigated strategies which help workers to avoid low priority work.

We present the basic design of the Bristol scheduler, discussing the data structures and the main algorithms. We also present performance results for the new scheduler using a number of benchmark programs and large applications. We show that the performance of the Bristol scheduler compares favourably with other schedulers. Our work also shows that special treatment of speculative work leads to improved performance.

On leave from (and present address) SZKI, IQSOFT, Donáti u. 35–45, Budapest, Hungary