CUDA-Accelerated Simulation of Brownian Dynamics
TimeTuesday, July 246:30pm - 8:30pm
DescriptionMolecular simulations are increasingly used to better understand biological and synthetic nanoscale systems. Brownian Dynamics is useful as a model for the interaction between molecules. An n-body problem, the calculation of pairwise forces between molecules is naively an O(n^2) algorithm. In our work, we begin with a single-threaded simulation of toluene molecules interacting with graphene. We reduce computational cost with verlet pairlists and cell decomposition, and to optimize cache performance for spatial locality, we explore space-filling curves as a means to sort the toluene molecules in 3-dimensional space into 1-dimensional memory. Finally, we accelerate the calculation of pairwise forces with CUDA.