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 10:50
EXASTAMP PISTON IMPACT SIMULATION Alizée Dubois and Thi. Piston directed compression and release on Tantalum (Vp = 1500 m.s 1) EAM potential (Ravelo et al, PRB 2013) t = 43,146 ps 83 Millions of atoms Particles Velocity 100 X 100 X 150 nm 3 4.8e+01 30 nodes with 64 cores per node 103680 h CPU 0.0e+00 18 Go / time step > 1000To in total 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.
    R. Ravelo et al. (2013) Shock-induced plasticity in tantalum single crystals: Interatomic potentials and large-scale molecular-dynamics simulations. Physical Review B
Share slide
Summary (AI generated)

The simulation involves 83 million atoms, and our objective is to analyze the resulting data effectively. We aim to detect voids, as well as determine their positions, growth velocities, and volume ratios. However, existing literature offers limited methodologies for this analysis.

One notable tool in the field is the VET code, which provides extensive post-processing capabilities. Despite its strengths, it is only parallelized using OpenMP, which constrains its performance based on the system architecture. As a result, when processing tens of millions of atoms, the tool becomes inefficient. Additionally, since VET is a post-processing tool, it requires the recording of vast amounts of data to achieve high temporal resolution, making it unsuitable for our specific analysis needs.