Avaliação do uso eficiente de algoritmos paralelos de filogenia em supercomputadores

Abstract

Bioinformatics, especially the areas of phylogeny and evolution of organisms, requires high performance computing (HPC) environments. The present work presents a behavioral analysis of the RAxML phylogeny tool in HPC environments. RAxML implements parallelism and task distribution by using threads and MPI libraries. Executions were performed in the Santos Dumont (SDumont) supercomputer. Efficiency results show the impact is generated by the variation in configuration, coupling and use of RAxML and SDumont. This information will allow a better understanding and efficient use of these specialized environments.