A consortium of MUMPS users
Last MUMPS User Days, June 1-2, 2017
Habilitation à diriger des recherches, September 2012 (slides)
- Sparse direct solvers
- High-performance computing
- MUMPS: MUltifrontal Massively Parallel Solvers for symmetric and unsymmetric complex/real sparse matrices
- MUMPS 5.1.2 (Oct. 2017) is
- Features include: out-of-core, efficient parallel processing of
symmetric indefinite matrices, time and memory optimizations,
orderings on compressed graphs, distributed Schur complement, reduced/condensed RHS,
multiple right hand-sides, sparse
multiple right-hand sides, distributed solution, determinant, forward elimination during
factorization, selected entries of the inverse, null pivot detection, parallel analysis.
MUMPS 5.1.2 (Oct. 2017)
(in pdf format)
The MUMPS Consortium is a consortium of industrial users who fund
the project in exchange of a number of services, see http://mumps-consortium.org
- Fédération Lyonnaise de Modélisation et Sciences Numériques
I am collaborating with the SEISCOPE consortium (LJK, Geoazur and ISTerre),
which focuses on wave propagation problems and seismic
imaging. The use of sparse direct solvers such as MUMPS is challenging
given the size of the problems involved. The SEISCOPE project is supported
by 13 industrial sponsors including major oil companies.
- ADT MUMPS, 3 years (2009-2012):
ADTMUMPS is an action of technological development funded
by Inria. This project provided support for 24 men-months of a young
engineer and part of time of a permanent engineer.
One goal of the project is to
improve daily work of MUMPS developers by improving the software
engineering aspects, and by reconsidering the validation tests and
tools to experiment the package.
- GRID TLSE (2002-):
This project was initially (2002-2005) funded by the ACI GRID. It consists of a Web portal
for expertise on sparse matrices, including software and databases, as well as bibliography and collections of sparse matrices.
Using the DIET middleware, users can submit requests of expertise for the solution of sparse
systems of linear equations. For example a typical request might be "which sparse matrix reordering heuristic leads to the
smallest number of operations for my matrix ?", or "which software is the most robust for this test problem ?"
The project is coordinated by ENSEEIHT-IRIT (Toulouse, France).
- SOLSTICE (ANR-06-CIS6-010, 2007-2010)
(SOLveurs et SimulaTIon en Calcul Extrême)
The objective of this project is to design and develop
high-performance parallel linear solvers that will be efficient to
solve complex multi-physics and multi-scale problems of very large
size (10 to 100 millions of equations). This project also comprises
LaBRI (coordinator), CERFACS, INPT-IRIT, CEA-CESTA, EADS-CCR,
EDF R&D, and CNRM. I am involved in tasks related
to out-of-core factorization and solution, parallelization of the
analysis phase of sparse direct solvers, rank detection, hybrid
direct-iterative methods and expertise site for sparse linear algebra.
funded project (2008-2009),
Scalable Sparse Linear Equation Solvers on Emerging Petascale Computers (CERFACS, Inria, INPT-IRIT,
Lawrence Berkeley National Lab).
- NSF-Inria Project(2001-2003): Robust parallel preconditioning methods:
bridging the gap between direct and iterative solvers.
- EGIDE-AURORA Project(2005): The impact of advanced reordering
techniques on state-of-the-art solution methods for sparse linear systems
of equations. With Jacko Koster (BCCS, PARALLAB, University of
Bergen). This project also involves Abdou Guermouche (PHD student at ENS Lyon),
as well as Patrick Amestoy (ENSEEIHT-IRIT, Toulouse,
France), and Stéphane Pralet and Luc Giraud (while at CERFACS, Toulouse, France).
Last Update: October 22, 2017