Missions

The CNRS is hiring a high-performance scientific computing engineer at the Lagrange laboratory. Scientific computing, particularly high-performance computing and high performance data analysis, has been at the heart of the Lagrange laboratory's scientific activities for many years. The world of scientific computing is currently undergoing major changes with the arrival of acceleration solutions. Multi-core architectures and, more recently, GPUs are now essential on machines dedicated to numerical computation (for example, in the French context, the CINES “AdAstra” machine is equipped with 338 AMD GPU nodes). More and more, numerical computing has to meet the challenge posed by the profusion of hardware architectures (NVIDIA and AMD GPUs, multi-core, fpga, etc.) and software (MPI, OpenMP, OpenACC, CUDA, etc.).
While the richness of this emerging environment offers a plethora of opportunities, it also comes with its share of challenges. In particular, it is now nearly impossible for developers to adapt their code to all the computing machine architectures available to them, often at considerable cost, and certainly not to maintain these adaptations over the long term.
In this context, several meta-libraries (e.g. KOKKOS) have emerged to enable developers to concentrate on high added-value tasks and build efficient, reusable and long-lasting software.
In this context, the high-performance scientific computing engineer to be recruited will be involved in porting code to GPUs. He/she will work in collaboration with several teams whose codes are operational on parallel machines but need to be adapted to modern architectures. He/she will design and develop the GPU porting of existing codes currently running on parallel CPU machines on national centres such as CINES, Idris, TGCC. He/she will also help the teams in the laboratory in the process of porting and using the code.

Activités

• Handling existing codes and their functioning on CPUs.
• Discussing with researchers and/or engineers to decide on the GPU porting strategy.
• Defining the steps to be carried out and the tests to be performed.
• Setting up compilation and test scripts.
• Porting the code to the GPU.
• Taking part in meetings with researchers to keep track of progress.
• Writing documentation relating to the porting of code to the GPU.

Compétences

• Programming languages (C or C++ or Fortran)
• Knowledge of scientific computing, MPI parallelization and GPU acceleration libraries (e.g. CUDA, Kokkos, oneAPI, SYCL).
• sbatch submission tools (e.g. Slurm)
• Command interpreters (bash/shell)
• Experience of version management software (e.g. git/svn),
• Knowledge of file transfer systems (e.g. sftp, scp, rsync, etc.)

Contexte de travail

The Lagrange laboratory (Observatoire Côte d'Azur, Université Côte d'Azur, CNRS) is located in Nice, France. It is a multidisciplinary laboratory that brings together teams in astrophysics (planetology, stellar and solar physics, galaxies and cosmology), fluid mechanics, signal and image processing and instrumentation for astronomical observation at high spatial resolution and high dynamics. Cross-disciplinary skills in high-performance computing are at the heart of the Laboratory' ability to develop new theories and models , and to compare them with observations acquired on large ground-based and space-based telescopes.
The successful candidate will work in close collaboration with the researchers and engineers who may be part of the research groups, and/or with the other engineers in the engineering and research group.

Working hours: full-time - teleworking possible under CNRS regulations.

Benefits: participation in health insurance, sustainable mobility package, social action (childcare, sport and leisure, ticketing, etc.).

The contract is for 12 months. There will probably be a possibility of extending the contract for a further 1-2 years.

Contraintes et risques

Working on several sites: Campus universitaire de Valrose, Observatoire de la Côte d'Azur, located about 10 km apart and both easily accessible by public transport.