BEAGLE 3 High-performance Computational Library for Phylogenetic Inference

International audienceMaximum-likelihood and Bayesian inference approaches to statistical phylogenetics require repeatedly computing of the observed sequence data likelihood. As increasingly cheaper sequencing technology provides phylogenomic studies with larger data set sizes, there stands a critical need to efficiently evaluate this likelihood function in order for phylogenetic computations to complete in reasonable time. The adoption of powerful computing architectures, in the form of multi-core central processing units and many-core graphics cards dedicated to scientific computing, offers unprecedented opportunity to perform massively parallel computation in many research fields, including likelihood evaluation in phylogenetics. In this chapter, we provide insight into the inner workings of BEAGLE, a high-performance likelihood-calculation platform for use on multi-core and many-core computer systems (ubiquitous nowadays in standard desktop computers and laptops) and available in several phylogenetic inference applications to improve computational performance