A Parallel Bottom-Up Resolution Algorithm Using Cilk

Rapid developments of multicore processors in the last ten years have accelerated the advancements in concurrency platforms. Performance of bottom-up resolution algorithms used in logic programming and artificial intelligent systems, can potentially be improved using the parallel programming constructs offered by these platforms (e.g., OpenMP, Cilk++, etc.). In this work we use Cilk++ to implement a parallel bottom-up resolution algorithm, and study how different parallel programming constructs affect its performance. Our experimental results show that a careful Cilk++ implementation of the algorithm can lead to significant speedup w.r.t. its traditional serial implementation.