Aceleração de autômatos celulares no contexto de biologia através de computação paralela em GPUS com OPENCL

• Ano de defesa: 2017
• Autor(a) principal: Pereira, Maelso Bruno Pacheco Nunes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Informática; Programa de Pós-Graduação em Informática; UFPB
• Palavras-chave em Português: Autômatos celulares; GPU; GPGPU; Histograma; PRNG; Gerador de números pseudo-aleatórios; Cellular automata; Histogram; Pseudo random number generator; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/12930

Cellular Automaton (CA) have its origins in the work of Von Neumann in the 40s and, since then, have become an important research topic with a wide range of applications, ranging from DNA sequencing to ecological dynamics. One aspect that may be of interest during a CA simulation is the evolution in the number of individuals of each species along time. This analysis can give important information about the dominance of certain species in a dynamical system, or identify aspects that might favor one or more species in detriment of others. CA simulations can be computationally very expensive tasks. Depending on the simulation domain size, number of dimensions or the number of individuals, these simulations can take several hours to complete. The evaluation of the number of individuals at each simulation time-step is an equally expensive task. Several acceleration techniques have been developed to improve the performance of CA simulations, and some of them take into account the evolution in the number of individuals along the simulation. In this work we propose an CA simulator which is capable of ef?ciently evaluate the evolution in the number of individuals of each species. High performance is obtained through the use of the massive parallelism of GPUs. The presented approach achieved a speed-up of 44 times when compared to a sequential implementation, and 26 times when compared to a traditional approach also in GPU.