On-the-fly clustering for exascale molecular dynamics simulations. - presented by Dr Alizée Dubois and Thierry Carrard

On-the-fly clustering for exascale molecular dynamics simulations.

Alizée Dubois and Thierry Carrard

ADThierry Carrard
Slide at 03:54
EXANBODY FRAMEWORK
Alizée Dubois and Thi..
N-BODY
exaDEM (Discrete Element Method) exaStamp (Molecular Dynamics)
exaSPH (Smooth Particule Hydrodynamic)
Carrard, T., et al. ExaNBody : A HPC Framework for N-Body Applications in Euro-Par 2023 : Parallel
Processing Workshops, (2024)
A. DUBOIS - T. CARRARD - COMPUTER PHYSICS COMMUNICATIONS SEMINAR SERIES - 03/03/25
II Votre écran est partagé par le biais de l'application app.zoom.us.
Arrêter le partage
Masquer
1
References
  • 1.
    T. Carrard et al. (2024) ExaNBody: A HPC Framework for N-Body Applications. Lecture Notes in Computer Science
Share slide
Summary (AI generated)

Hello everyone. I am Cheria, responsible for overseeing our computer science initiatives, particularly in simulation tools. Our journey began with the development of our Molecular Dynamics code, ExaMD, with a strong emphasis on high-performance computing facilities. Over the years, we have refined this tool and introduced numerous massively parallel algorithms, including GPU ports for most of these algorithms. This advancement has sparked interest in optimizing other simulation tools to leverage similar parallel algorithms tailored for specific simulations.

For example, we have created a specialized N-body simulation platform called Exabody. This platform serves as a foundation for any N-body based simulation code. Currently, we have three simulation codes built on Exabody that model various systems, including small grains of matter, rigid bodies like sand grains or powders, and complex rigid bodies that incorporate contact dynamics, friction, and collision detection.

In addition to our work with Molecular Dynamics, we are also exploring hydrodynamics through particle-based simulations using Smooth Particle Hydrodynamics. This approach is particularly advantageous for scenarios where traditional mesh-based methods are insufficient.