Contact and personnal information
|
Senior researcher at CNRS Associate professor at ÉTS Montréal Laboratoire de l'Informatique du Parallélisme École Normale Supérieure de Lyon 46 allée d'Italie, 69007 Lyon (previously at ILLS, at ÉTS Montréal) How to reach me:
|
|
Recent talks:
- University of Chicago (2025): Leveraging Expert Usage to Speed up LLM Inference with Expert Parallelism (slides)
- International Laboratory on Learning Systems, École de Technologie Supérieure de Montréal (2023) : Scheduling task graphs to reduce data movement (slides)
- Compas 2022, Amiens, Keynote "Parallélisme": Ordonnancement pour la mémoire et les transferts de données (slides [fr])
- Workshop "New Challenges in Scheduling Theory", Aussois (2022): Scheduling of Tasks Sharing Data on GPUs with limited memory (slides)
- Kickoff of the ANR Solharis project, Bordeaux (2020): Recent results and open questions on memory-aware DAG scheduling (slides)
PhD students:
Currently advised:
- Adrien Obrecht, co-supervised with Anne Benoit
- Félix Wirth, CIFRE thesis with the Kog company
- Maxime Gonthier, co-supervised with Samuel Thibault, now post-doc at Argonne Nat. Lab.
- Anthony Dugois, co-supervised with Louis-Claude Canon, now assistant professor at Université Marie & Louis Pasteur (formerly Université de Franche-Comté) in Besançon.
- Yishu Du, co-supervised with Yves Robert, on Fault-tolerant algorithms for iterative applications and batch schedulers.
- Changjiang Gou, co-supervised with Anne Benoit, on Communication- and memory-aware graph scheduling, now research engineer at Zhejiang Lab (China)
- Bertrand Simon, co-supervised with Frédéric Vivien, now a CNRS researcher at CC-IN2P3 (France)
- Julien Herrman, co-supervised with Yves Robert now a CNRS researcher at IRIT (France)
- Mathias Jacquelin, co-supervised with Yves Robert, now research engineer at Cerebras System (USA)
Funded Projects
I am part of the ANR SOLHARIS project (2019-2023), and responsible for the "Scheduling" work-package. This project extends the ANR SOLHAR project (2013-2018), which aimed at studying and designing algorithms and parallel programming models for implementing direct methods for the solution of sparse linear systems on emerging computers equipped with accelerators. In SOLHARIS, we are mainly interested in scaling the strategies developed during the SOLHAR project to distributed platforms.
Software
memDAG is library that handles task graphs (DAGs) with memory information. It is for example useful to compute or estimate the minimal sequential memory of a given DAG (or tree) of tasks, to compute the maximal memory of a parallel DAG processing and to tests several scheduling algorithms. It also includes an SP-ization algorithm, that is, an algorithm that transform any graph into a Series-Parallel graph by adding synchronisation vertices and edges.
|
Responsabilities and Service
Responsible for the "Commission des thèses" for the LIP Laboratory (2025--) Head of the ROMA research team (2020 -- 2023) Responsible of the computer science section of the competitive exam for entering ENS-Lyon (2017--2023) |
|
|
Teaching
In 2019 and 2020, I have given the course on Data Aware Algorithms of the "Master d'Informatique Fondamentale" at ENS Lyon. In 2008, 2010 and 2012, I was in charge of the course on Scheduling of the "Master d'Informatique Fondamentale" at ENS Lyon. During my PhD, I was responsible for the practicals of the course "Algo des Réseaux et de Telecoms". The material I made is still available here. |
|
Publications
The following publications are extracted from HAL and sould contain all my recent publications. The complete list of my publications is available on the DBLP Computer Science Bibliography. |
|