An Improved GPU Simulator for Spiking Neural P Systems

TitleAn Improved GPU Simulator for Spiking Neural P Systems
Publication TypeConference Contributions
Year of Publication2011
AuthorsCabarle, F., Adorna H., & Martínez-del-Amor M. A.
Conference NameSixth International Conference on Bio-Inspired Computing: Theories and Applications (BIC-TA)
ISBN Number978-1-4577-1092-6
Place PublishedPenang, Malaysia
Pages262 - 267
Date Published09/2011

Spiking Neural P (SNP) systems, variants of Psystems (under Membrane and Natural computing), are computing models that acquire abstraction and inspiration from the way neurons 'compute' or process information. Similar to other P system variants, SNP systems are Turing complete models that by nature compute non-deterministically and in a maximally parallel manner. P systems usually trade (often exponential) space for (polynomial to constant) time. Due to this nature, P system variants are currently limited to parallel simulations, and several variants have already been simulated in parallel devices. In this paper we present an improved SNP system simulator based on graphics processing units (GPUs). Among other reasons, current GPUs are architectured for massively parallel computations, thus making GPUs very suitable for SNP system simulation. The computing model, hardware/software considerations, and simulation algorithm are presented, as well as the comparisons of the CPU only and CPU-GPU based simulators.