Research methods
ACEMD: Accelerated molecular dynamics simulations in the microseconds timescale
The high arithmetic performance and intrinsic parallelism of recent graphical processing units (GPUs) can offer a technological edge for molecular dynamics simulations. ACEMD is a production-class bio-molecular dynamics (MD) simulation program designed specifically for GPUs which is able to achieve supercomputing scale performance of 40 nanoseconds/day for all-atom protein systems with over 23,000 atoms. We illustrate the characteristics of the code, its validation and performance. We also run a microsecond-long trajectory for an all-atom molecular system in explicit TIP3P water on a single workstation computer equipped with just 3 GPUs. This performance on cost effective hardware allows ACEMD to reach microsecond timescales routinely with important implications in terms of scientific applications. M. Harvey, G. Giupponi and G. De Fabritiis, ACEMD: Accelerated molecular dynamics simulations in the microseconds timescale, preprint (2009).
Performance of the Cell processor for biomolecular simulations
The Cell broadband engine is a new processor architecture created by Sony-Toshiba-IBM\cite{ibmcellsite} which allows for high computational performance and low production costs removing, by design, many important bottlenecks of standard processors. In the present version, it comprises one simplified power PC core (PPE) which runs the operating system and acts as a standard processor and 8 independent synergetic processing elements (SPEs). Main memory can be accessed only by the PPE core while each SPE can use its limited in-chip local memory (local store) accessed directly without any intermediate caching. This architectural design removes the memory bottleneck which is afflicting modern processors and furnishes a direct way to improve performance by adding more SPEs without having to rely only on clock frequency. Each core (PPE or SPEs) features a single instruction multiple data (SIMD) vector unit which gives a combined peak performance of around 230 Gflops at 3.2Ghz. I have looked at the important application case of molecular dynamics. G. De Fabritiis, Performance of the Cell processor for biomolecular simulations, Comp. Phys. Commun. 176, 670 (2007).pdf
Hybrid molecular dynamics (MD) - fluctuating hydrodynamics (FH)
Hybrid MD is a multiphysics (multiscale) coupled model between molecular dynamics (MD) and a continuum representation of a mesoscale fluid resolved by fluctuating hydrodynamics (FH). This very power technique allow simulation of large time and length scales keeping the molecular specificy in a sub-domain of the entire system (for instance a boundary condition). The two models MD and FH are separate executables which can be deploied on different computational resources but interface on-the-fly by exchanging momentum and mass. The coupling protocol requires to mantain a open MD system which is made possible by new algorithms of molecule insertion. G. De Fabritiis, R. Delgado-Buscalioni and P. V. Coveney, Modelling the mesoscale with molecular specificity, Phys. Rev. Lett. 97, 134501 (2006). pdf