Thierry Matthey

ProtoMol, an object-oriented framework for prototyping novel algorithms for molecular dynamics

ProtoMol is a high-performance framework in C++ for rapid prototyping of novel algorithms for molecular dynamics and related applications. Its flexibility is achieved primarily through the use of inheritance and design patterns (object-oriented programming). Performance is obtained by using templates that enable generation of efficient code for sections critical to performance (generic programming). The framework encapsulates important optimizations that can be used by developers, such as parallelism in the force computation. Its design is based on domain analysis of numerical integrators for molecular dynamics (MD) and of fast solvers for the force computation, particularly due to electrostatic interactions. Several new and efficient algorithms are implemented in ProtoMol. Finally, it is shown that ProtoMols sequential performance is excellent when compared to a leading MD program, and that it scales well for moderate number of processors. Binaries and source codes for Windows, Linux, Solaris, IRIX, HP-UX, and AIX platforms are available under open source license at http://protomol.sourceforge.net.

Parallel multigrid summation for the N-body problem

An @Q(n) parallel multigrid summation method (MG) for the N-body problem is presented. The method was originally devised for vacuum boundary conditions. Here, it is extended to periodic boundary conditions and implemented in parallel using force decomposition and MPI. MG is based on a hierarchical decomposition of computational kernels on multiple grids. For low accuracy calculations, appropriate for molecular dynamics, a sequential implementation is as fast or faster than particle mesh Ewald (PME). Our parallel implementation is more scalable than PME. The method can be combined with multiple time stepping integrators to produce a powerful simulation protocol for simulation of biological molecules and other materials. The parallel implementation is tested on both a Linux cluster with Myrinet interconnect and a shared memory computer. It is available as open-source at http://protomol.sourceforge.net. An auxiliary tool allows the automatic selection of optimal parameters for MG, and is available at http://mdsimaid.cse.nd.edu.

Overcoming Instabilities in Verlet-I/r-RESPA with the Mollified Impulse Method

The primary objective of this paper is to explain the derivation of symplectic mollified Verlet-I/r-RESPA (MOLLY) methods that overcome linear and nonlinear instabilities that arise as numerical artifacts in Verlet-I/r-RESPA. These methods allow for lengthening of the longest time step used in molecular dynamics (MD). We provide evidence that MOLLY methods can take a longest time step that is 50% greater than that of Verlet-I/r-RESPA, for a given drift, including no drift. A 350% increase in the timestep is possible using MOLLY with mild Langevin damping while still computing dynamic properties accurately. Furthermore, longer time steps also enhance the scalability of multiple time stepping integrators that use the popular Particle Mesh Ewald method for computing full electrostatics, since the parallel bottleneck of the fast Fourier transform associated with PME is invoked less often. An additional objective of this paper is to give sufficient implementation details for these mollified integrators, so that interested users may implement them into their MD codes, or use the program ProtoMol in which we have implemented these methods.

Dynamics of clustering in a binary Lennard-Jones material

We have performed constant temperature two-dimensional molecular dynamics simulations of a binary Lennard-Jones material, representing an idealized metallic alloy or a material containing a specified fraction of an additive atomic species. Differences in the interatomic potentials between the atomic species can lead to clustering of the alloy atoms. An exponential distribution of cluster size is obtained as time approaches infinity. For sufficiently strong attractions the distribution of cluster sizes becomes nearly independent of the force and the attained maximum cluster size reaches a saturation level.

Evaluation of MPI's One-Sided Communication Mechanism for Short-Range Molecular Dynamics on the Origin2000

In this paper we evaluate the possibilities of one-sided communication, a new feature of the MPI-2 standard, on the Origin2000 for relatively short-range molecular dynamics (MD) simulations. Our algorithm is based on an asynchronous message-passing multi-cell approach using MPI as message-passing layer and the Leap-Frog/Verlet algorithm for the time integration. We compare one-sided with two different two-sided communication approaches for typical production runs (105 - 109 atoms) where we discuss the communication vs. computation time for increasing number of processes. We also show how the partitioning of the problem affects the different communication approaches. Using one-sided communication we achieved 10-70% better performance over two-sided.

MDSIMAID: Automatic parameter optimization in fast electrostatic algorithms

MDSIMAID is a recommender system that optimizes parallel Particle Mesh Ewald (PME) and both sequential and parallel multigrid (MG) summation fast electrostatic solvers. MDSIMAID optimizes the running time or parallel scalability of these methods within a given error tolerance. MDSIMAID performs a run time constrained search on the parameter space of each method starting from semiempirical performance models. Recommended parameters are presented to the user. MDSIMAIDs optimization of MG leads to configurations that are up to 14 times faster or 17 times more accurate than published recommendations. Optimization of PME can improve its parallel scalability, making it run twice as fast in parallel in our tests. MDSIMAID and its Python source code are accessible through a Web portal located at http://mdsimaid.cse.nd.edu.

ProtoMol: a molecular dynamics research framework for algorithmic development

This paper describes the design and evaluation of ProtoMol, a high performance object-oriented software framework for molecular dynamics (MD). The main objective of the framework is to provide an efficient implementation that is extensible and allows the prototyping of novel algorithms. This is achieved through a combination of generic and object-oriented programming techniques and a domain specific language. The program reuses design patterns without sacrificing performance. Parallelization using MPI is allowed in an incremental fashion. To show the flexibility of the design, several fast electrostatics (N-body) methods have been implemented and tested in ProtoMol. In particular, we show that an O(N) multi-grid method for N-body problems is faster than particle-mesh Ewald (PME) for N > 8,000. The method works in periodic and nonperiodic boundary conditions. Good parallel efficiency of the multi-grid method is demonstrated on an IBM p690 Regatta Turbo with up to 20 processors for systems with N = 102, 104 and 106. Binaries and source code are available free of charge at http://www.nd.edu/~lcls/protomol.

A Parallel Split Operator Method for the Time Dependent Schrödinger Equation

We describe the parallelization of a spectral method for solving the time dependent Schrodinger equation in spherical coordinates. Two different implementation of the necessary communication have been implemented and experiments on a real application are presented. The excellent runtime and accuracy figures demonstrate that our approach is very efficient. With the very encouraging speed-up, we claim that the numerical scheme is parallelizable and our MPI-implementation efficient.

Molecular dynamics simulations of sliding friction in a dense granular material

Friction in a dense two-dimensional granular material is studied by molecular dynamics simulations. Initially the particles are ordered in a triangular lattice where the top layer feels a constant vertical pressure and horizontal velocity field. The dynamics of the material at low velocities is found to be characterized by avalanche-type motion during short time intervals followed by long stationary periods of internal stress buildup. At low external pressures, the time-averaged shear stress is found to be nearly constant for low values of the normal pressure and proportional to the load for higher pressures.

Ion Coulomb Crystals With Long Range Order in RF-traps

In this contribution, we present results on the structural properties of single‐component as well as two‐component ion Coulomb crystals confined in a linear Paul trap. Metastable long range ordered structures as the face‐centered cubic (fcc) and body‐centered cubic (bcc) lattices are observed in single‐component ion crystals of sizes down to ∼ 1000 ions. In the case of two‐component ion crystals we observe a stable long range ordered structure in one of the trapped components, which is compared with a Molecular Dynamics (MD) simulation of a two‐component crystal.