Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2019 Nov 1;68(6):1052-1061.
doi: 10.1093/sysbio/syz020.

BEAGLE 3: Improved Performance, Scaling, and Usability for a High-Performance Computing Library for Statistical Phylogenetics

Daniel L Ayres  1 , Michael P Cummings  1 , Guy Baele  2 , Aaron E Darling  3 , Paul O Lewis  4 , David L Swofford  5 , John P Huelsenbeck  6 , Philippe Lemey  2 , Andrew Rambaut  7   8 , Marc A Suchard  9   10   11
Affiliations

BEAGLE 3: Improved Performance, Scaling, and Usability for a High-Performance Computing Library for Statistical Phylogenetics

Daniel L Ayres et al. Syst Biol. .

Abstract

BEAGLE is a high-performance likelihood-calculation library for phylogenetic inference. The BEAGLE library defines a simple, but flexible, application programming interface (API), and includes a collection of efficient implementations for calculation under a variety of evolutionary models on different hardware devices. The library has been integrated into recent versions of popular phylogenetics software packages including BEAST and MrBayes and has been widely used across a diverse range of evolutionary studies. Here, we present BEAGLE 3 with new parallel implementations, increased performance for challenging data sets, improved scalability, and better usability. We have added new OpenCL and central processing unit-threaded implementations to the library, allowing the effective utilization of a wider range of modern hardware. Further, we have extended the API and library to support concurrent computation of independent partial likelihood arrays, for increased performance of nucleotide-model analyses with greater flexibility of data partitioning. For better scalability and usability, we have improved how phylogenetic software packages use BEAGLE in multi-GPU (graphics processing unit) and cluster environments, and introduced an automated method to select the fastest device given the data set, evolutionary model, and hardware. For application developers who wish to integrate the library, we also have developed an online tutorial. To evaluate the effect of the improvements, we ran a variety of benchmarks on state-of-the-art hardware. For a partitioned exemplar analysis, we observe run-time performance improvements as high as 5.9-fold over our previous GPU implementation. BEAGLE 3 is free, open-source software licensed under the Lesser GPL and available at https://beagle-dev.github.io.

Keywords: Bayesian phylogenetics; GPU; maximum likelihood; multicore processing; parallel computing.

PubMed Disclaimer

Figures

Figure 3.
Figure 3.
Relative performance gains (fold-speedup) on two systems for analysis with BEAST v1.10.5 and BEAGLE v3.1.2 for an unpartitioned codon analysis of the previously described dengue virus data set (see Performance Evaluation section), comprising 3330 codons. For the BEAGLE GPU implementation, this data set requires 21.5 GB of memory, and thus we scale computation by pattern block across two library instances, each running on a separate GPU. Benchmarks are for 10,000 iterations, and the single-threaded, non-vectorized, version of BEAGLE CPU running on System 1 is used as a reference.
Figure 1.
Figure 1.
Relative performance gains (fold-speedup) for a challenging highly partitioned nucleotide-model analysis using various combinations of implementations and versions of the BEAGLE library, and hardware resources, with BEAST and with MrBayes. We report fold-speedup on the log-scale, relative to the total run time when using the native double-precision likelihood calculator on the slowest system (denoted with an asterisk) for each program.
Figure 2.
Figure 2.
Log–log contour plot depicting BEAGLE-GPU memory usage for BEAST and MrBayes nucleotide-model analyses with four rate categories and double precision floating-point arithmetic, over a range of problem sizes in terms of number of OTUs and of unique site patterns. The amount of memory depicted as values below the dashed isolines convey the upper boundary for the memory size indicated.
Figure 5.
Figure 5.
Absolute (throughput in billions of partial likelihood calculations per second) and relative (fold-speedup relative to the slowest performance observed at any number of unique site patterns) performance scaling with problem size for implementations of the BEAGLE library Version 3.1.2 and the Phylogenetics Likelihood library Version 2 on nodes of the Comet Supercomputer available via CIPRES. The data are simulated nucleotide sequences for a tree of 128 OTUs.
Figure 4.
Figure 4.
Relative performance gains (fold-speedup) for analysis with MrBayes v3.2.7 and BEAGLE v3.1.2, demonstrating the scalability across a range of MPI processes and hardware devices on nodes of the Comet Supercomputer available via CIPRES. Results denoted with a formula image were benchmarked on nodes with a newer-specification CPU and GPU (the CPU executed non-BEAGLE code). As reference, we use the double-precision MrBayes calculator, running as a single CPU process. We use the same dengue virus data set as before (see Performance Evaluation section), with an increase in the number of Markov chains to four and replicate runs to two.
See this image and copyright information in PMC

References

    1. Altekar G., Dwarkadas S., Huelsenbeck J.P., Ronquist F.. 2004. Parallel Metropolis coupled Markov chain Monte Carlo for Bayesian phylogenetic inference. Bioinformatics. 20:407–415. - PubMed
    1. Ayres D.L., Cummings M.P.. 2017a. Configuring concurrent computation of phylogenetic partial likelihoods: accelerating analyses using the BEAGLE library In: Ibrahim S., Choo K., Yan Z., Pedrycz W., editors. Algorithms and architectures for parallel processing. Vol. 10393 of Lect. Notes Comput. Sc. Helsinki, Finland: ICA3PP; 2017. p. 533–547.
    1. Ayres D.L., Cummings M.P.. 2017b. Heterogeneous hardware support in BEAGLE, a high-performance computing library for statistical phylogenetics. 2017 46th International Conference on Parallel Processing Workshops (ICPPW), Bristol, 2017 p. 23–32.
    1. Ayres D.L., Cummings M.P.. 2018. Rerooting trees increases opportunities for concurrent computation and results in markedly improved performance for phylogenetic inference. 2018 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW). Vancouver, BC, 2018. p. 247–256.
    1. Ayres D.L., Darling A., Zwickl D.J., Beerli P., Holder M.T., Lewis P.O., Huelsenbeck J.P., Ronquist F., Swofford D.L., Cummings M.P., Rambaut A., Suchard M.A.. 2012. BEAGLE: an application programming interface and high-performance computing library for statistical phylogenetics. Syst. Biol. 61:170–173. - PMC - PubMed

Publication types