Transparent Parallelization of Constraint Programming

The availability of commodity multi-core and multi-processor machines and the inherent parallelism in constraint programming search offer significant opportunities for constraint programming. They also present a fundamental challenge: how to exploit parallelism transparently to speed up constraint programs. This paper shows how to parallelize constraint programs transparently without changes to the code. The main technical idea consists of automatically lifting a sequential exploration strategy into its parallel counterpart, allowing workers to share and steal subproblems. Experimental results show that the parallel implementation may produces significant speedups on multi-core machines.