1. Onika: a Component-Based HPC Platform for Numerical Simulation

Onika (Object Network Interface for Knit Applications) is a component-based HPC software platform for building numerical simulation codes.

Onika is the foundation of the exaNBody particle simulation platform, but it is not tied to N-Body problems, nor to any other specific application domain. Existing applications built on Onika’s components include Molecular Dynamics, particle-based fluid simulations using methods such as Smoothed Particle Hydrodynamics (SPH), and rigid body simulations using methods such as the Discrete Element Method (DEM).

See Motivation for why Onika is built around a component-based architecture.

1.1. Key Characteristics

  • Language: Implemented in C++20.

  • Build system: CMake.

  • Configuration: YAML for user input files.

  • Parallelization: MPI and OpenMP for distributed and shared-memory parallelism.

  • GPU acceleration: CUDA and HIP.

1.2. Onika in other codes

Onika is used by several codes:

  • ExaNBody, for N-Body simulations (open source) [3].

  • ExaStamp, for Molecular Dynamics simulations (open source) [4].

  • ExaDEM, for Discrete Element Method simulations (open source) [5]. ExaDEM combines exaNBody for its HPC performance layer with the physical modeling capabilities of Rockable [6], notably its R-shape particle representation.

  • HippoLBM, for Lattice Boltzmann Method simulations (open source). HippoLBM reuses computational kernels from an earlier code, DEMLBM3D, integrated into Onika operators, with data structures reworked for GPU execution and a domain decomposition algorithm added to support MPI+GPU parallelism.

  • ExaSPH, for Smoothed Particle Hydrodynamics simulations (not open source) and Coupling ExaSPH with EuroPlexus via onika (not open source) [1].

  • ExaCoLD, for LBM-DEM-IBM coupling (soon to be open source).

  • A prototype of CEA’s PLEIADES-HPC platform (not open source) [2].

References

[1]

Alberto Beccantini, Fabienne Bliard, Pascal Bouda, Stanislas de Lambert, Florence Drui, Pascal Galon, Olivier Jamond, and Nicolas Lelong. EUROPLEXUS: un code de référence pour la dynamique rapide et l'interaction fluide-structure. In CSMA 2022-15ème Colloque National en Calcul des Structures. 2022.

[2]

Stéphane Bernaud, Isabelle Ramiere, Guillaume Latu, and Bruno Michel. PLEIADES: a numerical framework dedicated to the multiphysics and multiscale nuclear fuel behaviour simulation. Annals of Nuclear Energy, 205:110577, 2024. doi:10.1016/j.anucene.2024.110577.

[3]

Thierry Carrard, Raphaël Prat, Guillaume Latu, Killian Babilotte, Paul Lafourcade, Lhassan Amarsid, and Laurent Soulard. ExaNBody: a HPC framework for N-Body applications. In Euro-Par 2023: Parallel Processing Workshops, 342–354. Cham, 2024. Springer Nature Switzerland. doi:10.1007/978-3-031-50684-0_27.

[4]

Emmanuel Cieren, Laurent Colombet, Samuel Pitoiset, and Raymond Namyst. Exastamp: a parallel framework for molecular dynamics on heterogeneous clusters. In Euro-Par 2014: Parallel Processing Workshops: Euro-Par 2014 International Workshops, Porto, Portugal, August 25-26, 2014, Revised Selected Papers, Part II 20, 121–132. Springer, 2014. doi:10.1007/978-3-319-14313-2_11.

[5]

Raphaël Prat, Thierry Carrard, Lhassan Amarsid, Vincent Richefeu, Carlo-Elia Doncecchi, Paul Lafourcade, Guillaume Latu, and Jean-Mathieu Vanson. ExaDEM: a HPC application based on exaNBody targeting scalable DEM simulations with complex particle shapes. Journal of Open Source Software, 10(106):7484, 2025. doi:10.21105/joss.07484.

[6]

Vincent Richefeu, Gaël Combe, Lhassan Amarsid, Raphaël Prat, Jean-Mathieu Vanson, Saied Nezamabadi, Patrick Mutabaruka, Jean-Yves Delenne, and Farhang Radjaï. Advanced strategies for discrete simulations with three-dimensional R-shapes in Rockable framework. Computer Physics Communications, 320:109997, 2025. doi:10.1016/j.cpc.2025.109997.