Teraflops and beyond: GPU-based MD exploration of emergent phenomena

Molecular dynamics (MD) simulation of emergent phenomena can be computationally demanding because of the broad range of length and time scales that must be covered, ranging from the individual particles out to where the collective behavior is expressed; the fact that simulations of this type are often subject to unpredictable outcomes is a further complication. Examples of MD studies of emergent behavior include discrete-particle modeling of hydrodynamic instabilities (e.g., thermal convection cells), complex segregation processes in granular systems modeled with inelastic particles (e.g., in a rotating drum), and supramolecular self-assembly (e.g., the growth of icosahedral shells corresponding to viral capsids). The comparatively large and long simulations required for these problems benefit substantially from massively parallel GPU-based implementation, with even a single GPU typically providing an order of magnitude speedup over a conventional CPU. A sampling of newly obtained exploratory results for these and similar problems [arXiv:1409.5958] will be described, along with the methodology; the results offer a tantalizing hint of the kinds of phenomena that can be explored, and what might be achieved given the appropriate resources.