Experimenting with X10 for Parallel Constraint-Based Local Search

TL;DR

This paper explores using the PGAS language X10 to develop a parallel Constraint-Based Local Search solver, demonstrating near-linear speed-ups across various benchmarks due to effective parallelization strategies.

Contribution

It presents the first implementation of a Constraint-Based Local Search solver in X10, highlighting its architectural advantages and parallel performance benefits.

Findings

Near-linear speed-ups achieved on benchmarks

Effective parallelization of the Adaptive Search algorithm

X10's suitability for parallel constraint solving

Abstract

In this study, we have investigated the adequacy of the PGAS parallel language X10 to implement a Constraint-Based Local Search solver. We decided to code in this language to benefit from the ease of use and architectural independence from parallel resources which it offers. We present the implementation strategy, in search of different sources of parallelism in the context of an implementation of the Adaptive Search algorithm. We extensively discuss the algorithm and its implementation. The performance evaluation on a representative set of benchmarks shows close to linear speed-ups, in all the problems treated.