A new hybrid parallel algorithm for MrBayes

Abstract

MrBayes, a popular program for Bayesian inference of phylogeny, has not been fast enough for Biologists when dealing with large real-world data sets. This paper presents a new parallel algorithm that combines the chain-partitioned parallel algorithm with the chain-parallel algorithm to obtain higher concurrency. We test the proposed hybrid algorithm with the two old algorithms on a heterogeneous cluster. The results show that, the hybrid algorithm actually converts more CPU cores into higher speedup compared with the two control algorithms for all of four real-world DNA data sets, therefore is more practical. © Springer-Verlag Berlin Heidelberg 2010.