Scheduling speculative work in MUSE and performance results

Abstract

Work which may later be pruned is called speculative work. In this paper we present and evaluate a simple and efficient strategy, used in the Muse OR-parallel Prolog system, for better scheduling of speculative work. The strategy concentrates workers on the leftmost available work in the Prolog tree as'long as there exists enough parallelism, thus emulating the sequential Prolog execution as much as possible. This strategy therefore makes it less probable that unnecessary work is executed. A new cut scheme that reduces the amount of speculative work is also presented. The performance results of our strategy are compared with the performance results of similar strategies implemented in the Aurora OR-parallel Prolog system. The comparison shows that our strategy performs quite well.

Parallel execution of prolog programs

“…An issue that arises in the presence of pruning operators such as cuts and commits during or-parallel execution is that of speculative work [Hausman 1989;Ali and Karlsson 1992b;Beaumont and Warren 1993;Sindaha 1992]. Consider the following program:…”

“…Essentially, such a scheduler should bias all the workers to pick work that is within the scope of a cut from branches to the left in the corresponding subtree rather than from branches to the right [Ali and Karlsson 1992b;Beaumont 1991;Beaumont and Warren 1993;Sindaha 1992].…”

“…The MUSE or-parallel Prolog system has been designed and implemented on a number of UMA and NUMA computers (Sequent Symmetry, Sun Galaxy, BBN Butterfly II, etc.) [Ali and Karlsson 1990b;1990a;Ali et al 1992;Ali and Karlsson 1992b;Karlsson 1992]. It supports the full Prolog language and programs run on it with almost no user annotations.…”

“…Extensions are carefully added to preserve the high efficiency of SICStus leading to a negligible overhead which is significantly lower than in other or-parallel models. The MUSE scheduler supports efficient scheduling of speculative work and nonspeculative work [Ali and Karlsson 1992b]. For purposes of scheduling, the Prolog tree is divided into two sections: the right section contains voluntarily suspended work and the left section contains active work.…”

High Level Thread-Based Competitive Or-Parallelism in Logtalk

“…This form of work is known as speculative work [14]. An advanced scheduler must be able to reduce to a minimum the speculative computations and at the same time maintain the granularity of the work scheduled for execution [15,16].…”