Saving time and energy with oversubscription and semi‐direct Møller–Plesset second order perturbation methods. Issue 11 (30th April 2017)

• Record Type: Journal Article
• Authors: Fought, Ellie L.; Sundriyal, Vaibhav; Sosonkina, Masha; Windus, Theresa L.
• Is Part Of: Journal of computational chemistry. Volume 38:Issue 11(2017)
• Page Start: 830
• Page End: 841
• Publication Date: 2017-04-30
• Subjects: energy savings -- oversubscription -- semi‐direct algorithms -- MP2 -- multithreading; Chemistry -- Data processing -- Periodicals; 542.85
• DOI: 10.1002/jcc.24756
• Languages: English
• ISSNs: 0192-8651

Abstract

Abstract : In this work, the effect of oversubscription is evaluated, via calling 2 n, 3 n, or 4 n processes for n physical cores, on semi‐direct MP2 energy and gradient calculations and RI‐MP2 energy calculations with the cc‐pVTZ basis using NWChem. Results indicate that on both Intel and AMD platforms, oversubscription reduces total time to solution on average for semi‐direct MP2 energy calculations by 25–45% and reduces total energy consumed by the CPU and DRAM on average by 10–15% on the Intel platform. Semi‐direct gradient time to solution is shortened on average by 8–15% and energy consumption is decreased by 5–10%. Linear regression analysis shows a strong correlation between time to solution and total energy consumed. Oversubscribing during RI‐MP2 calculations results in performance degradations of 30–50% at the 4 n level. © 2017 Wiley Periodicals, Inc. Abstract : A straightforward method of oversubscription is used in conjunction with semi‐direct Møller–Plesset second order perturbation (MP2) calculations. Oversubscription involves calling more virtual computer processes than physical processors. When oversubscribing the time to solution for both semi‐direct MP2 energy and semi‐direct MP2 gradient calculations is shortened significantly and the total energy used by the computer to complete the calculation is reduced. Oversubscribing aids the processes in overcoming the I/O bottleneck that arises in many semi‐direct methods.