This thesis presents a framework for highly parallel and hybrid molecular dynamics simulations. One of the key attributes is its modularity, which eases the exchange of different components. The framwork therefore allows a quick installation of new molecule models and the development and comparison of different load-balancing strategies, which are the core area of the thesis. From the application point of view, a special focus is set on nucleation processes, which are characterized by very heterogeneous particle distributions. Different algorithmic strategies, based on diffusion, space-filling curves, k-d-trees and graphs, are optimised to tackle the challenges given by this application from the field of chemical engineering.
«
This thesis presents a framework for highly parallel and hybrid molecular dynamics simulations. One of the key attributes is its modularity, which eases the exchange of different components. The framwork therefore allows a quick installation of new molecule models and the development and comparison of different load-balancing strategies, which are the core area of the thesis. From the application point of view, a special focus is set on nucleation processes, which are characterized by very heter...
»