GAIDAMOUR Jérémie

Fonction : Ingénieur de recherche CNRS
Équipe : Équations aux dérivées partielles
Domaines de recherche :
  • Simulation of quantum mechanics problems (Schrödinger and Gross-Pitaevskii equations)
  • Numerical methods for partial differential equations
  • Parallel solvers for large sparse linear systems (hybrid direct-iterative methods, multigrid)
Mots clés :
  • High Performance Computing
  • Bose-Einstein condensation
  • Pseudo-spectral methods
  • Domain decomposition methods
  • Preconditioners
  • Sparse linear algebra
Coordonnées :

IECL – Site de Nancy
Faculté des sciences et Technologies
Campus, Boulevard des Aiguillettes
54506 Vandœuvre-lès-Nancy

Email : jeremie.gaidamour@univ-lorraine.fr
Téléphone : 03 72 74 57 78
Bureau : 202 b

I am a CNRS Research Engineer in Scientific Computing and HPC at the Institut Élie Cartan de Lorraine, the Mathematics Research Laboratory of the University of Lorraine.

I’m interested in high-order numerical methods for computing the stationary states and dynamics of Bose-Einstein Condensates (BEC) modeled by Gross-Pitaevskii Equations (GPEs). I’m currently working on the parallel implementation of numerical methods to simulate large scale 2D and 3D realistic problems in quantum physics.

Previously, I worked on parallel linear solver. In particular, I developped an hybrid direct-iterative method based on a Schur complement domain decomposition during my PhD and an AMG solver based on energy minimization approaches during my post-doc. I also worked on the support team at IDRIS, one of the three national centers for High-Performance Computing in France. I also spent two years with the Grid’5000 team.

Here is a list of my software development projects:

  • BEC2HPC : parallel spectral methods for computing the stationary states and dynamics of Bose-Einstein Condensates (BEC) modeled by Gross-Pitaevskii Equations (GPEs) (Python, C++, MPI).
  • MueLu : a parallel multigrid solver based on smoothed aggregation and energy minimization approaches (C++, MPI, part of the Trilinos project).
  • MueMat : a Matlab toolbox to experiment with multigrid preconditioners.
  • Xpetra : a lightweight linear algebra interface (Python, C++) for Trilinos.
  • HIPS : a parallel hybrid direct-iterative sparse solver . It uses a domain decomposition where the Schur complement is solved using a Krylov method preconditioned by a global incomplete factorization (a supernodal ILUK) (C, MPI).

PhD Thesis

Journal articles

Research Reports

  • Jérémie Gaidamour, Dimitri Lecas, Pierre-François Lavallée. Introducing OpenMP Tasks into the HYDRO Benchmark. PRACE White Papers, 2014.
  • Jérémie Gaidamour, Pascal Hénon. Etude de performance des méthodes hybrides directes/itératives du logiciel HIPS sur des cas d’électromagnétisme du CEA. Rapport de fin de contrat, 2008.

International conferences

  • Jérémie Gaidamour, Xavier Antoine, A parallel framework for the numerical simulation of Bose–Einstein condensates, International Conference on Scientific Computation and Differential Equations (SciCADE), Innsbruck, Austria, 2019.
  • Jérémie Gaidamour, Jonathan J. Hu, Christopher M. Siefert and Ray S. Tuminaro. MueMat : a MATLAB Toolbox to Experiment with New Multigrid Preconditioners. International Congress on Industrial and Applied Mathematics (ICIAM), Vancouver, Canada, 2011.
  • Jérémie Gaidamour, Jonathan J. Hu, Christopher M. Siefert et Ray S. Tuminaro. Algebraic Multigrid using Energy-Minimization : a general framework to develop intergrid transfer operator. In proceedings of the International Conference on Preconditioning Techniques for Large Sparse Matrix Problems in Scientific and Industrial Applications (PRECOND), Bordeaux, France, 2011.
  • Jérémie Gaidamour, Jonathan J. Hu, Christopher M. Siefert et Ray S. Tuminaro. MueLu : Designing an extensible framework for multigrid algorithm research. Copper Mountain Conference on Multigrid Methods, Copper Mountain, USA, 2011.
  • Jérémie Gaidamour, Jonathan J. Hu, Christopher M. Siefert et Ray S. Tuminaro. A General AMG Strategy for Addressing the Near Null-space Based on Energy Minimization. International Conference on Domain Decomposition Methods (DD), San Diego, USA, 2011.
  • Pascal Hénon et Jérémie Gaidamour. Algorithms to build robust hybrid direct/iterative sparse linear solvers. International Conference on Preconditioning Techniques for Large Sparse Matrix Problems in Scientific and Industrial Applications, Hong Kong, 2009.
  • Jérémie Gaidamour et Pascal Hénon. A parallel direct/iterative solver based on a Schur complement approach. 11th IEEE International Conference on Computational Science and Engineering, pages 98–105, Sao Paulo, Brazil, 2008.
  • Jérémie Gaidamour et Pascal Hénon. HIPS : a parallel hybrid direct/iterative solver based on a Schur complement approach. PMAA’08, Neuchâtel, Suisse, 2008.
  • Jérémie Gaidamour, Pascal Hénon, Jean Roman, et Yousef Saad. Parallel resolution of sparse linear systems by mixing direct and iterative methods. International Symposium on Iterative Methods in Scientific Computing (IMACS), Lille, France, 2008.
  • Jérémie Gaidamour, Pascal Hénon, Jean Roman, et Yousef Saad. An hybrid direct-iterative solver based on a hierarchical interface decomposition. International Conference on Preconditioning Techniques for Large Sparse Matrix Problems in Scientific and Industrial Applications, Toulouse, France, 2007.

International Workshops

  • Jérémie Gaidamour, Xavier Antoine, A parallel computation of the stationary states of rotating Bose-Einstein condensates using an iterative pseudo-spectral method, Workshop on Mathematical and Computational Methods for Quantum Systems, Montreal, Canada, 2018.
  • Jérémie Gaidamour et Pascal Hénon. Comparison of algorithms to build an efficient schur complement preconditioner in hips. Sparse Days, CERFACS, Toulouse France, 2009.
  • Jérémie Gaidamour et Pascal Hénon. HIPS : a parallel hybrid direct/iterative solver based on a Schur complement approach. Sparse Days, CERFACS, Workshop de VECPAR’08, Toulouse France, 2008.
  • Jérémie Gaidamour, Pascal Hénon, Jean Roman et Yousef Saad. An hybrid direct-iterative solver based on the Schur complement approach. 8th Workshop of the ERCIM Working group, Salerne Italie, 2006.

Posters

  • Jérémie Gaidamour, Jonathan J. Hu, Christopher M. Siefert et Ray S. Tuminaro. MueLu : Designing a New Multigrid Solver for the Trilinos Project. Supercomputing Conference (SC), New Orleans, USA, 2010.

User Guides

A selection of tutorials I have given in the past: