Solips, parallel computing serving industry
In order to accelerate digital simulations - which are sometimes very time-consuming - numerical analysts are turning to parallel computing. This process, which is becoming increasingly widespread in industry, is at the heart of the work carried out by Cédric Lachat, computer scientist and co-founder of the Solips project with François Pellegrini.
Can you tell us about the aim of your research?
Cédric Lachat : The research focuses on digital simulation and parallel computing. The initial idea was to find a way to accelerate a part of the computing chain that was still sequential ; for example, the analysis of the air flow around an aircraft. This mobilises various software programs - some of which will work in parallel - for several hours and even several days. My project aims at "parallelising" a tool that is still sequential in this computing chain. In order to do so a re-meshing - i.e. a modification to the initial partitioning, or mesh - of the part to be simulated is often required, in order to guarantee the simulation is of sufficient quality. Our software makes re-meshing with several processing units - in other words parallel re-meshing - possible.
How does your solution work?
Cédric Lachat : We have designed a set of algorithmic methods that enable the distribution of data over several processing units and allow us to obtain a level of quality equivalent to that of sequential software. This means savings in time, efficiency and also costs. I particularly rely on the PaMPA software library, work on which began in 2009 with a thesis supervised by François Pellegrini and Cécile Dobrzynski from the TADAAM and CARDAMOM research teams at Inria Bordeaux. This library makes it possible to efficiently manipulate distributed meshes on a large number of processors. PaMPA offers reliable and constant functionalities that maintain the simulations' effectiveness.
What led you to work in this field?
Cédric Lachat : The subject is quite innovative and has great potential today. Although pioneering work had already been done on parallel re-meshing, what was missing was a method that could use existing software - which explains my interest in this slightly unusual venture. It is still quite a specialised niche market, but I am optimistic about its growth potential, especially in the aeronautics and automobile fields.
What are the next steps for the project?
Cédric Lachat : My intention is to validate the effectiveness of this software in the industrial domain. The first tests have already been carried out with Dassault Aviation. In parallel, I intend to carry out market research in order to identify potential clients. Finally I would like to create my company, either during 2016 or early 2017. In that capacity, I am being helped by the Aquitaine regional business incubator. Moreover, thanks to the grant obtained through the i-LAB competition, I intend to undertake training in order to be able to make this project a success.
When we look at your background, we notice that nothing suggested that you might one day become an engineer and company director...
Cédric Lachat : That's true, I have quite an unusual background. After my A levels, I studied at medical school for two years, before joining a university course in computer science, then in engineering. I think that the work placement with François Pellegrini, researcher at the Bordeaux computer science research laboratory and at Inria, was both significant and decisive in the direction my studies were going to take. Since the end of my thesis, I have been devoting all my time to this PaMPA software. It's a bit like my baby, and I want to see it grow up.
These articles could interest you:
For more details
The start-up project
- Name : Solips
- Date of creation : 2016-2017
- Location : Bordeaux
- Collaborators : 2
- Domain : parallel computing