Description du projet
La prochaine génération de logiciels de mécanique des fluides numérique pourrait passer à l’exascale
Les méthodes de Lattice Boltzmann (LBM) sont une classe de méthodes de mécanique des fluides numérique (MFN) qui constituent une alternative fiable à la MFN classique. Dans de nombreuses applications d’ingénierie, elles se sont révélées plus rapides d’un ordre de grandeur que les approches Navier-Stokes dans des scénarios comparables. Les LBM sont également bien adaptées à l’exploitation des architectures avancées des superordinateurs grâce à la vectorisation, aux accélérateurs matériels et aux ordinateurs massivement parallèles. Le projet SCALABLE, financé par l’UE, réunira des partenaires industriels et universitaires de premier plan en vue d’atteindre des performances, une évolutivité et une efficacité énergétique sans précédent pour un logiciel industriel de MFN basé sur les LBM.
Objectif
In SCALABLE, eminent industrials and academic partners will team up to achieve the scaling to unprecedented performance, scalability, and energy efficiency of an industrial LBM-based computational fluid dynamics (CFD) software.
Lattice Boltzmann methods (LBM) have already evolved to become trustworthy alternatives to conventional CFD. In several engineering applications they are shown to be roughly an order of magnitude faster than Navier-Stokes approaches in a fair comparison and in comparable scenarios.
In the context of EuroHPC, LBM is especially well suited to exploit advanced supercomputer architectures through vectorization, accelerators, and massive parallelization.
In the public domain research code waLBerla, superb performance and unlimited scalability has been demonstrated, reaching more than a trillion (10^12) lattice cells already on Petascale systems. waLBerla performance excels because of its uncompromising unique, architecture-specific automatic generation of optimized compute kernels, together with carefully designed parallel data structures. waLBerla, however, is not compliant with industrial applications due to lack of a geometry engine and user friendliness for non-HPC experts.
On the other hand, the industrial CFD software LaBS already has such industrial capabilities at a proven high level of maturity, but it still has performance worthy of improvement. Therefore, SCALABLE will transfer the leading edge performance technology from waLBerla to LaBS, thus breaking the silos between the scientific computing world and physical flow modelling world to deliver improved efficiency and scalability for LaBS to be prepared for the upcoming European Exascale systems.
The project outcomes will directly benefit to the European industry as confirmed by the active involvement of Renault & Airbus in the project and will additionally contribute to fundamental research.
Champ scientifique
- natural sciencescomputer and information sciencessoftware
- natural sciencescomputer and information sciencescomputational science
- natural sciencesphysical sciencesclassical mechanicsfluid mechanicsfluid dynamicscomputational fluid dynamics
- natural sciencesmathematicspure mathematicsgeometry
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringcomputer hardwaresupercomputers
Programme(s)
Régime de financement
IA - Innovation actionCoordinateur
92350 Le Plessis Robinson
France