Abstract
dc:description.abstractA range of specialized Molecular Dynamics (MD) methods have been developed in order to overcome the challenge of reaching longer timescales in systems that evolve through sequences of rare events. In this talk, we consider Parallel Trajectory Splicing (ParSplice) which works by generating large number of MD trajectory segments in parallel in such a way that they can later be assembled into a single statistically correct state-to-state trajectory, enabling parallel speedups up to N, the number of parallel workers. The prospect of strong-scaling MD is extremely enticing given the continuously increasing scale of available computational resources: on current peta-scale platforms N can be in the hundreds of thousands, which opens the door to MD-accurate millisecond-long atomistic simulations; extending such a capability into the exascale era could be transformative.In practice, however, the ability for ParSplice to scale increasingly relies on predicting where the trajectory will be found in the future. With this insight in mind, we develop a maximum likelihood transition model that is updated on the fly and make use of an uncertainty-driven estimator to approximate the optimal distribution of trajectory segments to be generated next. In addition, we investigate resource optimization schemes designed to fully utilize computational resources in order to generate the maximum expected throughput.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy (PhD)
- Level thesis:degree_level
- Dissertation
- Discipline thesis:degree_discipline
- Physics and Astronomy
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Garmon, Andrew
- Contributors dc:contributor
-
- Murray Daw
- Jian He
- Brad Meyer
- Jens Oberheide
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://open.clemson.edu/all_dissertations/2716