I am member of the LIHPC research lab and director of research at CEA, France. I work on designing and providing meshing algorithms and software for the purpose of numerical simulations. My main research interests are about: quadrilateral and hexahedral mesh generation, mesh adaptation and mesh partitioning.
We produce quadrilateral and hexahedral mesh generation for the purpose of hydrodynamics and areodynamics codes. In this context, preferred meshes are said block-structured. It is my main research topic. Today, I consider that frame-field based global parameterizations, including Polycube, are the option to follow. As a complement, with researchers of IBISC, we investigate how to update block structures using reinforment learning techniques.
Generating meshes for numerical simulation codes raises the question of getting the “right” mesh. Trying to anwser leads to investigate how to generate/adapt a mesh considering data provided by the simulation code itself. It means to adapt the mesh to numerical fields and not only to a geometrical shape.
The last component of my reseach is mesh partitioning. It is a key feature for running HPC simulation codes efficiently. Our specificity is that we partition meshes and not graphs and we deal both with architecture constraints (MPI process, CPU threads, GPU threads) and code requirements (multi-stages physics, load balancing on the fly).
Accreditation to supervise research, 2013
Université de Poitiers, France
PhD in Computer Science, 2002
Université d'Evry Val d'Essonne, France
Master degree in Mathematics, 1999
Université d'Evry Val d'Essonne, France
I was born the 30th of April 1975. I defended my Ph.D. in 2002 on how to use algebraic specifications to study and define geometric operations, directed by Pascale Le Gall and Yves Bertrand. I was hired by CEA in 2004 as a researcher-enginner to work on the subject of quadrilateral and hexahedral meshing.
Since 2009, I’m Associate professor in Computer Science at the University of Évry-Val d’Essonne, University of Paris-Saclay. I’ve mainly taught software engineering (scrum, designs patterns, testing, etc.), component and object oriented design and meshing.
In 2015-2016, I was invited scientitist at Lawrence Livermore National Laboratories, CA, USA, to design new algoritmhs for generation hexahedral-dominant meshes (partially funded by the French defense procurement agency, or DGA). Since 2018, I lead a CEA group of engineers and researcher that develop, maintain and provide user support for HPC tools and libraries: CAD & meshing, scientific visualization, physics databases (equations of state, opacity,. . . ). I am now, since january 2022, director of research at CEA and CEA fellow and expect to transfer research solutions into robust software.
Publications are given by year with the following indexation: [JI] for articles in international journal with selection committee, [JN] for articles in French journal with selection committee, [O] for articles in book, [CI] for invited presentation in international and national conferences, [CP] for presentations in international conferences with selection committee and proceedings, [CS] for presentations in international conferences with selection committee.
N. Le Goff, F. Ledoux and J.-Ch. Janodet, Overlay Grid Mesh Adaptation using Discrete Interface Reconstruction, in 13th Symp. on Trends in Unstructured Mesh Generation, 16th USCNM Conference, Chicago, USA, July 2021. [CS]
N. Le Goff, F. Ledoux and J.-Ch. Janodet, Intercode Hexahedral Meshing from Eulerian to Lagrangian Simulations, chapter book in SEMA-SIMAI Springer Series in honour of the 60th birthday of Oubay Hassan, to appear in 2021. [O]
F. Ledoux Maillage structuré pour le calcul scientifique : Et si la solution venait de l’informatique graphique? In JFIG 2020, Journées Française d’Informatique Graphique, LORIA, Nancy, 26 Novembre 2020. [CI]
S. Morais, C. Chevalier, and F. Ledoux. A Multilevel Mesh Partitioning Algorithm Driven by Memory Constraints. SIAM Workshop on Combinatorial Scientific Computing (CSC), Feb. 11–13, Seattle, WS, U.S.A., 2020. [CP]
GMDS is a set of C++ libraries that provides data structures and algorithms (services) to represent, generate and modify 2D and 3D meshes. It mainly focuses on the generation of quadrilateral and hexahedral structured meshes for purposes in numerical simulation.
Most of the PhD students that work in our team use GMDS to develop their own works and contributions. GMDS is avalaible under LGPL licence on Github. See
Coupe is a modular, multi-threaded partitioning library. It implements a variety of algorithms that can be used to partition meshes, graphs and numbers (see the API docs for a list). These algorithms can be composed together to build relevant partitions of your data. See
MGX is our 3D software used to interactively create and handle 3D BRep shape and block structures. It will be released soon under Affero GPL licence.