Intel^{®} Parallel Computing Center @ RWTH Aachen
Project
Molecular dynamics is a highly computeintensive simulation tool for determining the thermodynamic, kinetic, and transport properties of various materials. The rise of affordable multicore workstations and parallel computing clusters has led to the widespread proliferation of molecular dynamics as a tool through almost all areas of science and engineering, with applications as diverse as protein folding and aggregation to nanoparticle rheology in industrial processing to materials for nuclear waste storage and carbon capture.
In spite of the increasing availability of various accelerator hardware, molecular dynamics codes have been not kept pace, thereby preventing codes from taking full advantage of the latest generation of hardware, including the Intel^{®} Xeon Phi™ line of processors. The goal of our Intel^{®} Parallel Computing Center is to optimize some of the most important computational kernels in the LAMMPS molecular dynamics package for Intel^{®} architectures.
One area which we will address in our work is the problem of multibody potentials, which go beyond the standard pairwise potentials normally used and incorporate threebody, fourbody, and sometimes manybody interactions as well. Two such methods that are frequently used in the literature are the AIREBO and Tersoff potentials, which are encountered in simulations of carbon nanotubes, graphene, and other hydrocarbons. By changing how neighbor lists are stored, and by rearranging the data used by the calculations, we will better exploit the intranode parallelism and vectorization capabilities of the Intel^{®} Xeon Phi™ coprocessors.
The most critical part of our work will focus on the longrange solvers present in LAMMPS. In many largescale molecular simulations, these solvers, which are reponsible for calculating the forces resulting from either electrostatic or dispersion forces, represent the largest computational bottleneck, often accounting for as much as 90 to 95 percent of the total computational expense of a simulation. Consequently, any efficiency gains achieved in these algorithms can have a major impact on the MD community at large. We will exploit the builtin vectorization capabilities of the Intel^{®} Xeon Phi™ by adjusting how data is packed into arrays that handle particle mapping as well as the Poisson solver routines in the particleparticle particlemesh (PPPM) algorithms in LAMMPS. We will also explore possible improvements in the API which connects LAMMPS to the underlying Fast Fourier transform solvers that drive the PPPM algorithm.
Principal Investigators
Prof. Paolo Bientinesi, Ph.D.
Paolo Bientinesi is Professor for AlgorithmOriented Code Generation for HighPerformance Architectures in the Computer Science department at RWTH Aachen University; he leads the group HighPerformance and Automatic Computing within the Aachen Institute for Computational Engineering Science (AICES).
Prof. Bientinesi studied at the University of Pisa (MS, 1998) and at The University of Texas at Austin (Ph.D., 2006); he was a postdoctoral associate at Duke University prior to joining RWTH Aachen. His expertize lies in performance modeling, numerical linear algebra, and automation. He often collaborates with the European Commission as expert evaluator, and is the recipient of the 2009 Karl Arnold Prize from the North RhineWestphalian Academy of Sciences and Humanities.
Prof. Ahmed E. Ismail, Ph.D.
Prof. Ahmed E. Ismail leads the Molecular Simulations and Transformations research group, which is affiliated with the TailorMade Fuels from Biomass Cluster of Excellence, the Aachener Verfahrenstechnik, and the AICES Graduate School at RWTH Aachen University. Prof. Ismail received his bachelor's and doctorate in chemical engineering from Yale and MIT; after graduation, he was a postdoc and later technical staff member at Sandia National Laboratories.
His research group performs largescale molecular dynamics simulations for a number of applications, including biomass dissolution, interfacial structure and dynamics, and connections to coarsegrained methods
Articles & Talks

IPCC @ RWTH Aachen: Optimization of multibody and longrange solvers in LAMMPS
IPCC Showcase, Nov. 2016

The Vectorization of the Tersoff MultiBody Potential: An Exercise in Performance Portability
Slides from the presentation at SC'16, Nov. 2016

Accelerating ParticleParticle ParticleMesh Methods for Molecular Dynamics
Slides from Intel PCC 2016 Fall Forum, Oct. 2016

The Vectorization of the Tersoff MultiBody Potential: An Exercise in Performance Portability
arXiv preprint, SC'16, 2016

IPCC @ RWTH Aachen University, Showcase: KNL results
Slides from talk at Intel booth during ISC'16, 2016

Molecular Dynamics Research Enhanced by More Scalable LAMMPS HPC Code
Article in Scientific Computing, 2016

Optimization of multibody and longrange solvers in LAMMPS
Slides from Intel PCC EMEA Meeting in Ostrava, 2016

IPCC @ RWTH Aachen: Optimization of multibody and longrange solvers in LAMMPS
Slides from project showcase, 2016

The Tersoff manybody potential: Sustainable performance through vectorization
Article in Proceedings of SC'15 workshop: Producing High Performance and Sustainable Software for Molecular Simulations, 2015

The Tersoff manybody potential: Sustainable performance through vectorization
Slides from SC'15 workshop: Producing High Performance and Sustainable Software for Molecular Simulations, 2015
Code on Github
Repository of code for the Tersoff potential's optimization 
Repository of code for the Buckingham potential's optimization 
Additional Material
Team
Rodrigo Canales  
Markus Höhnerbach  
William McDoniel 
Links
 RWTH Aachen
 AICES  Aachen Institute for Advanced Study in Computational Engineering Science
 Molecular Simulations and Transformations
 LAMMPS
 Intel^{®} Parallel Computing Centers