A comparison between parallelization approaches in molecular dynamics simulations on GPUs

• Published in: Journal of computational chemistry Vol. 36; no. 1; pp. 1 - 8
• Main Authors: Rovigatti, Lorenzo, Šulc, Petr, Reguly, István Z., Romano, Flavio
• Format: Journal Article
• Language: English
• Published: United States Blackwell Publishing Ltd 05.01.2015; Wiley Subscription Services, Inc
• Subjects: Central processing units; Comparative analysis; CPUs; graphics processing unit; Mathematical analysis; Molecular chemistry; molecular dynamics; parallelization; Performance evaluation; Simulation; soft matter; graphics processing unit; soft matter; molecular dynamics; parallelization
• ISSN: 0192-8651; 1096-987X
• DOI: 10.1002/jcc.23763

We test the relative performances of two different approaches to the computation of forces for molecular dynamics simulations on graphics processing units. A “vertex‐based” approach, where a computing thread is started per particle, is compared to an “edge‐based” approach, where a thread is started per each potentially non‐zero interaction. We find that the former is more efficient for systems with many simple interactions per particle while the latter is more efficient if the system has more complicated interactions or fewer of them. By comparing computation times on more and less recent graphics processing unit technology, we predict that, if the current trend of increasing the number of processing cores—as opposed to their computing power—remains, the “edge‐based” approach will gradually become the most efficient choice in an increasing number of cases. © 2014 Wiley Periodicals, Inc. A comparison between two parallelization approaches to be used in molecular dynamics simulations on GPUs was performed. A more aggressive edge‐based approach, where a thread is started per interaction, is compared to a more standard vertex‐based approach where a thread is started per each particle. Three different potential interactions are tested. If the trend of increasing the number of computing units on GPUs is continued, the edge‐based approach will become the best choice in an increasing number of cases.