Webinar Series: Samuel Newcome

Towards Algorithm Selection for High-Performance Discrete Element Method Simulations

An introduction to AutoPas and its relevance for DEM.
News
Events
Webinar
DEM
Performance
Authors
Affiliations

JP Morrissey

Kevin Hanley

Catherine O’Sullivan

Published

October 30, 2026

Event Details

Event Date: Tuesday, November 17th, 14:00.
Location: Online.

Join via Teams

Towards Algorithm Selection for High-Performance Discrete Element Method Simulations

Speaker: Samuel Newcome

Affiliation: School of Computation, Information and Technology; Technical University of Munich

Samuel obtained a Bachelor’s and Master’s in Mathematical Sciences and Modern Applications in Mathematics, respectively, at the University of Bath. Since 2022, he has been working as scientific staff in the group of Prof. Bungartz at the Chair of Scientific Computing in Computer Science at the Technical University of Munich; officially a doctoral candidate since 2024.

Abstract

In short-range N-body simulations, such as the Discrete Element Method (DEM) and Molecular Dynamics (MD), the majority of a simulation’s computational cost typically comes from contact detection and force calculation between particle pairs.  As such, many algorithms have been developed for efficient, parallelised contact detection and force calculation. 

In MD, it has been found that the optimal algorithm is simulation-dependent, and an optimal algorithm for one simulation could perform an order of magnitude or more slower than the optimal algorithm for another simulation. The optimal algorithm can also change during a simulation and across different spatial regions of a simulation.  This led to the development of AutoPas, a particle simulation backend that aims to select and configure the optimal algorithm at runtime from an extensive algorithm library.

In this talk, we will introduce the motivation for AutoPas, some of its algorithms, its methods for algorithm selection, and more recent work extending its algorithm library for polydisperse DEM simulation. 

Want to Volunteer?

Will you be our next presenter? We are always on the look-out for interesting topics that may be of interest to community members. If you think you could contribute to the community by delivering a talk that is not solely research-focused, please get in contact with us.

Please direct any questions related to this or any offers to present to the CCC-ParaSolS mailbox: parasols@ed.ac.uk

Reminder: Join the Community

Also, don’t forget to join the CCC-ParaSolS community if you have not already done so!

Back to top