| Date: Friday, January 14, 2011
Title: Utilizing GPU's for Fast and Flexible Many-Particle Dynamics Simulations
Abstract: HOOMD-blue is an open source code for performing molecular dynamics and related many-body dynamics simulations on graphics processing units (GPUs). Current GPUs are capable of teraflops of floating-point computation and hundreds of gigabytes per second of memory bandwidth. High demand from the gaming industry and healthy competition among the manufacturers makes GPUs widely available and inexpensive. All calculations in HOOMD-blue, including the neighbor list build, pair force computation, integration, et cetera are performed on the GPU to attain the maximum performance possible. In typical benchmarks, HOOMD-blue on a current generation GPU is about 80-100 times faster than on a single CPU core. This efficient execution is achieved without any lack of generality and a wide variety of capabilities are present in the code, including standard bond, pair, angle, dihedral and improper potentials, along with the common NPT, NVE, NVT, and Brownian dynamics integration routines. All of this functionality is easy to configure via an extremely simple job scripting language. The simulation code is object-oriented, well documented, and easy to modify. In this talk, I present an overview of HOOMD-blue and its capabilities, discuss the GPU algorithms that drive it, and demonstrate the performance that it delivers. Special attention is given to the advantages offered by the latest generation of GPU hardware (GF100).
Speaker: Joshua Anderson
Institution: University of Michigan
|
