On domain decomposition with space filling curves for the parallel solution of the coupled Maxwell/Vlasov equations

Christian Konrad 1
1 NACHOS - Numerical modeling and high performance computing for evolution problems in complex domains and heterogeneous media
CRISAM - Inria Sophia Antipolis - Méditerranée , JAD - Laboratoire Jean Alexandre Dieudonné : UMR6621
Abstract : Space filling Curves (SFCs) are increasingly used for combinatorial scientific computing and in particular for designing fast domain decomposition (partitioning) methods. In the context of parallel particle simulations for solving the system of Maxwell/Vlasov equations with a coupled FE/PIC (Finite Element/Particle-In-Cell) unstructured mesh based solver, one has to deal with a two-constraint partitioning problem. Moreover, this problem has to be solved several times during the simulation. Therefore, a fast and scalable partitioning problem is required. For this purpose, we propose here a new SFC based method which is well adapted to multi-constraint partitioning problems. This method is compared to graph based partitioning methods from the widely used MeTiS tool. Experimental results show that the proposed SFC based method is at least 100 times faster than MeTiS to the disadvantage of edge-cuts that are between 2 to 4 times worse than those achieved by the MeTiS methods.
Complete list of metadatas

Cited literature [40 references]  Display  Hide  Download

https://hal.inria.fr/inria-00331382
Contributor : Christian Konrad <>
Submitted on : Thursday, October 16, 2008 - 2:43:57 PM
Last modification on : Thursday, May 3, 2018 - 1:32:55 PM
Long-term archiving on : Tuesday, October 9, 2012 - 1:50:49 PM

File

RR-6693.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00331382, version 1

Citation

Christian Konrad. On domain decomposition with space filling curves for the parallel solution of the coupled Maxwell/Vlasov equations. [Research Report] RR-6693, INRIA. 2008. ⟨inria-00331382⟩

Share

Metrics

Record views

243

Files downloads

604