| Membrane computing, known as P systems, is abstracted from the structure andfunctioning of living cell as well as interactions of cells in tissue and organ. P systems are aclass of distributed parallel computing models. From the membrane structure, P systemshave three types: cell-like P systems, tissue-like P systems and neural-like P systems.Parallel computing feature is one of the advantages of the P system, which is rather attractivefor solving a wide range of application problems. But now, the parallel computing of Psystems can’t be simulated really because of the computer’s serial structure.GPU (Graphic Processing Unit) is a concept relative to the CPU. Initially, it is in order toassist the CPU to process the image. It has parallel hardware architecture and powerfulfloating point capability, can achieve the hardware acceleration for image processing. How tosimulate parallel computing power of the P systems is a hot research topic in membranecomputing area. The emergence of the GPU, specifically its support of parallel computing ofmatrix operations, provide a new way for the simulation of parallel computing of P systems.In this paper, two kinds of spiking neural P systems are considered and we propose thematrix representations of their parallel computing and give the corresponding implementationalgorithms with GPU. In this paper we have two research works as follow:(1) The matrix representation of spiking neural P system with exhaustive use of rules isproposed. Based on the matrix representation, a GPU implementation algorithm ofthe spiking neural P system is developed. The simulation results on several casestudies illustrate the feasibility of GPU implementation of spiking neural P systemwith exhaustive use of rules.(2) For spiking neural P systems with delay, the matrix representation of their parallelcomputing is proposed. Then its GPU algorithm is developed based on the matrixrepresentation. The feasibility and effectiveness of the GPU implementation ofspiking neural P system with delay are verified on several case studies. |
PDF Full Text Request