Rigid Body Constraints In Molecular Dynamics Simulations Optimized for Graphics Processing Units

Monday, October 17, 2011: 10:40 AM
Conrad B (Hilton Minneapolis)
Trung D. Nguyen1, Carolyn L. Phillips2, Joshua A. Anderson1 and Sharon C. Glotzer3, (1)Department of Chemical Engineering, University of Michigan, Ann Arbor, MI, (2)Applied Physics, University of Michigan, Ann Arbor, MI, (3)Chemical Engineering and Biomedical Engineering, University of Michigan

We have designed and implemented massively parallel rigid body constraint algorithms in HOOMD-Blue[1], an open-source general purpose molecular dynamics simulation package optimized for graphics processing units (GPUs). In the HOOMD-Blue code package, rigid constraints can be used seamlessly with non-rigid parts of the system (e.g., point particles, bonded particles and walls), and with an array of integration methods (e.g. NVE, NVT, NPT, and Brownian Dynamics). We have also incorporated the FIRE energy minimizer[2], reformulated to be applicable to mixed systems of rigid bodies and non-rigid particles. For our typical simulations, the GPU implementation running on a single NVIDIA® GTX 480 card is twice as fast as LAMMPS[3] running on 32 CPUs.
[1]    HOOMD-Blue, http://codeblue.umich.edu/hoomd-blue/.
[2]    E. Bitzek, P. Koskinen, F. Gaehler, M. Moseler, P. Gumbsch, Structural relaxation made simple, Physical Review Letters 2006, 97 170201.
[3]    LAMMPS, http://lammps.sandia.gov/.

Extended Abstract: File Not Uploaded