GPU accelerated implementation of NCI calculations using promolecular density. Issue 14 (25th March 2017)

• Record Type: Journal Article
• Authors: Rubez, Gaëtan; Etancelin, Jean‐Matthieu; Vigouroux, Xavier; Krajecki, Michael; Boisson, Jean‐Charles; Hénon, Eric
• Is Part Of: Journal of computational chemistry. Volume 38:Issue 14(2017)
• Page Start: 1071
• Page End: 1083
• Publication Date: 2017-03-25
• Subjects: graphics processing unit -- noncovalent interactions -- high performance computing -- CUDA -- electron density -- NCI; Chemistry -- Data processing -- Periodicals; 542.85
• DOI: 10.1002/jcc.24786
• Languages: English
• ISSNs: 0192-8651

Abstract

Abstract : The NCI approach is a modern tool to reveal chemical noncovalent interactions. It is particularly attractive to describe ligand–protein binding. A custom implementation for NCI using promolecular density is presented. It is designed to leverage the computational power of NVIDIA graphics processing unit (GPU) accelerators through the CUDA programming model. The code performances of three versions are examined on a test set of 144 systems. NCI calculations are particularly well suited to the GPU architecture, which reduces drastically the computational time. On a single compute node, the dual‐GPU version leads to a 39‐fold improvement for the biggest instance compared to the optimal OpenMP parallel run (C code, icc compiler) with 16 CPU cores. Energy consumption measurements carried out on both CPU and GPU NCI tests show that the GPU approach provides substantial energy savings. © 2017 Wiley Periodicals, Inc. Abstract : Molecular interactions (noncovalent interactions [NCI]) are forces, either attractive or repulsive, between molecules. They are involved in important processes like boiling or crystallization or drug action. The NCI methodology provides a visual picture of these interactions from grid‐based calculations relying on the electron density knowledge. A graphics processing unit (GPU) accelerated NCI algorithm is described that leads to a 39‐fold speedup compared to an OpenMP parallel run with 16 CPU cores. The NCI GPU implementation is attractive in terms of runtime and energy efficiency.