Scalable N-Queens Solving on GPGPUs via Interwarp Collaborations

In this paper we present how recent hardware revisions and newly introduced approaches to thread collaboration in NVIDIA GPUs and the CUDA toolkit can be used to design an extensible, scalable GPU-based solver for the N-Queens problem. We discuss various design choices ranging from memory structure, to low-level optimisations on newer GPU hardware that result in strong performance when solving the N-Queens problem using an optimised solving algorithm that can be applied to other similar in nature problems.