Conducted byGuillaume Latu
DescriptionParticle-In-Cell (PIC) codes have become an essential tool for the numerical simulation of many physical phenomena involving charged particles, in particular beam physics, space and laboratory plasmas including fusion plasmas. Genuinely kinetic phenomena can be modeled by the Vlasov-Maxwell equations which are discretized by a PIC method coupled to a Maxwell field solver. Actual physics problems involving the self-consistent motion of relativistic particles in truly 3D complex geometries can now be envisioned on today’s massively parallel supercomputers. Our aim in our project is to develop new PIC solvers based on high order Maxwell solvers coupled with a relativistic particle pusher. In this project, we will develop massively parallel self consistent Particle-In-Cell Finite Element Time Domain solvers that can handle large and geometrically complex structures. A final part of the project will be to port the code to the Grid and assess its efficiency. One of our objectives in the Grid research field is to federate big computational resources located on two or three sites. The goal is to use efficiently hundreds of processors to compute a high-performance simulation on the Grid. That means to provide a good load balancing strategy; consequently, we need a detailed model of the application behavior in term of algorithmic complexities, and of computation and communication costs. Assessing the global application performance implies to derive a model able to predict the performance of an end-to-end single connection in presence of cross-traffic between sites. Because this problem is very hard, we will only consider Grids whose nodes have substantial computation capabilities with a high-bandwidth network; we let aside the discussion concerning desktop Grids, i.e. Grids composed of numerous PCs operated by individuals and located on various networks. The target platform will primarily be GRID 5000.
- Nodes involved: 10
- Sites involved: 1
Shared by: Guillaume Latu
Last update: 2008-11-20 10:21:35