Parallel discrete molecular dynamics simulation with speculation and in-order commitment

Abstract

Discrete molecular dynamics simulation (DMD) uses simplified and discretized models enabling simulations to advance by event rather than by timestep. DMD is an instance of discrete event simulation and so is difficult to scale: even in this multi-core era, all reported DMD codes are serial. In this paper we discuss the inherent difficulties of scaling DMD and present our method of parallelizing DMD through event-based decomposition. Our method is microarchitecture inspired: speculative processing of events exposes parallelism, while in-order commitment ensures correctness. We analyze the potential of this parallelization method for shared-memory multiprocessors. Achieving scalability required extensive experimentation with scheduling and synchronization methods to mitigate serialization. The speed-up achieved for a variety of system sizes and complexities is nearly 6× on an 8-core and over 9× on a 12-core processor. We present and verify analytical models that account for the achieved performance as a function of available concurrency and architectural limitations.

Applications of Discrete Molecular Dynamics in biology and medicine

“…The implicit solvent used with DMD provides an opportunity for parameterization to take into account the crowded cellular milieu, and the event-driven nature of the algorithms that advance the simulations are more appropriate to large, crowded systems than are continuous potentials with necessary integration over increasingly complex energetic potentials. In the near future, the combination of parallelization [7,33] and increasing processing power will likely make DMD the method of choice for the first atomistic cellular simulations and beyond.…”

Discrete Molecular Dynamics: An Efficient And Versatile Simulation Method For Fine Protein Characterization

“…Therefore many of the earlier collision predictions will remain valid if the participating atoms k and l are located sufficiently far from both i and j and the k–l collision takes place within a short time period after the i–j collision. The feasibility of the event-based parallelization approach has been recently demonstrated by Khan and Herbordt 48 using a scalable implementation on up to 8 CPUs in shared-memory system.…”

“…The parallelization approach described in Khan and Herbordt 48 splits the DMD simulation cycles into several stages. First, every collision event is predicted based on the current atoms positions and velocities.…”

Efficient event-driven simulations of hard spheres

“…If the event scheduled for particle i is invalidated by a collision between two other particles, the event is typically converted into an update event, which simply rechecks for future collisions and neighborhood events without changing the particle trajectory. Although this means that sometimes the same collision is predicted more than once, the reduction in computation cost due to scheduling fewer events can result in significant speedups [58]. An alternative strategy, used in e.g.…”

“…Nonetheless, several effective methods for parallelization of EDMD simulations have been demonstrated (see e.g. [58,64]), and can in principle be extended to the EDMD codes presented here.…”