"Molecular dynamics using graphics processors and CUDA technology"
Boyarchenkov A.S., Potashnikov S.I.

A parallel implementation of interparticle interaction computation in molecular dynamics with isolated boundary conditions on graphics processors using NVIDIA CUDA is considered. An efficient CUDA version using Newton's third law is first proposed. Full code optimization details are given. On NVIDIA GeForce 8800 GTX video card, a remarkable speed-up is achieved: up to 660 times in the case of 49152 particles compared to the scalar code running on AMD Athlon64 2 GHz CPU.

Keywords: molecular dynamics, parallel computing, graphics processing unit, CUDA technology

Boyarchenkov A.S.     e-mail: boyarchenkov@gmail.com
Potashnikov S.I.     e-mail: potashnikov@gmail.com