The path we try to follow this 2021 year for COOP
COOP focuses on transverse activities towards improving, optimizing and refactoring scientific codes at a sustainable pace. In 2021, we declined this into the four following axes.
Bringing scientific tools to Safran Group, Release “Hanami” (march) and “Ice-Tea” (sept.).
“Hanami” (march) and “Ice-Tea” (sept.) are our bi-annual release of software to the Safran group. This is about the industrialization of our research tools -mainly combustion and CFD- towards a shareholder.
While we do care a lot about our shareholders satisfaction, there are two non-profit advantages for Cerfacs: The industrialization activity of COOP is, first of all, a research catalyst for research teams. The cycle is the following:
-
Upgrade a scientific tool, out of a research project to an engineering design tool, industrial proof.
-
Support the engineers in their day-to-day use of the tools.
-
Collect real-life experience and new situations
-
Build a new research project to tackle the new situation.
At the same time, industrialization is a recurrent full stack work ranging from the early selection of a technical solution to the final support on the installed tool. By giving us this first hand experience, it allows COOP-CSE to develop a unique expertise on the transfer of scientific tools to industry and collaborators. Industrialization, is therefore a the source of many trainings on transverse skills related to scientific tools creation. For example, the work for Safran on the AVTP solver gave us the experience to write guide to work with Fortran Namelists.
-
Read more on the industrialization history with the history of Opentea releases (VPN required).
-
A major innovation in 2021 is
Lemmings, an open-source helper to deliver customizable workflows for industrial use on clusters. Read more on the introduction to Lemmings. -
The second main innovation is the
Tekigowrapper for mesh adaptation. Read more on the introduction to Tekigo. -
We clarified the AVBP versioning in use for the occasion, read about the Road and Scouts semantic versioning strategy.
Creating tools for Exascale production
Facilitating the exploitation of HPC resources, and preparing for the future of computations are recurrent topics in Centers of Excellence. These are also transverse concerns for our Shareholders. Here is a list of our creations in this direction.
-
Runcrawleris an open-source utility to profile run production on a cluster. We are currently woking to release it to PyPI. It was born from the Excellerat CoE. Read more in our blog post or in the excellerat whitepaper. -
Flinteris an open source utility to evaluate the compliance of a Fortran codebase to a custom standard. It was born from the Excellerat CoE and is available on PyPI. See what insight you can get on our dedicated post -
Lemmingsis an open source utility to help the creation and transfer of HPC workflows. It was born from the Excellerat CoE, refined thanks to our collaborators at Safran Aircraft, and is now used for the COEC CoE. As soon as our customers feel the product is satisfactory, we will release it on PiPY. You can read more about it on our dedicated post
Acquiring and evaluating new Softwares
We investigate softwares born outside Cerfacs that could prove useful for our research teams.
Concerning applied maths, we try to investigate the popular Finite Elements frameworks and compare these softwares to our in-house solvers. We started in 2021 with Fenics. Read more about:
- a 3D Fenics Thermal parallel solver on complex non-structured meshes (comparable to AVTP) written in few lines.
- a 3D Fenics Helmholtz solver on complex non-structured meshes (comparable to AVSP, without the parallel aspect yet) written in few lines. In Nov. 2021 a Phd from Cerfacs CFD team is interested in a 2D curvilinear version for is thesis work.
- a simple incompressible 2D Fenics Navier Stokes solver was built to meet the need of a Cerfacs Global Change Phd for a simple CFD solver in the context of data assimilation.
This tech watch on Fenics eventually opened two transverse collaborations proving the initiative was worth the effort. FireDrake and FreeFem could be under focus next year if opportunities arise.
Concerning HPC numerical frameworks, we will evaluate Friedrich Alexander Universitat’s waLBerla as a numerical testbed for Lattice Boltzmann equations resolution using code generation.
- The SCALABLE will -among other goals- test waLBerla and proLB on the same benchmarks. (See Scalable benchmark configurations).
Diffusion of technical experience
As Training is a major transverse goal for CERFACS, we try in COOP to develop new ways to diffuse the knowledge accumulated at CERFACS.
- The Gitlab.com CERFACS’s group is mirroring several of our open source projects. A fair share of these project have their documentation automatically build on ReadTheDocs (e.g. opentea documentation) and, if these are Python packages, you can find them on PyPI (e.g. h5nav, a command-line .hdf explorer ).
- The blog you are currently reading was created in Sept. 2020 with Pelican. This is our team knowledge base -keeping us together even in remote- but also a nice way to share a lot of info. This is a very sustainable form of asynchronous learning.
- Some concepts are too complex to explain via a static blog post. Within the frame of EuroCC National Competence Center, we are exploring dynamic technical blog posts based on Idyll. Read our Work in Progress on the Cerfacs explorable Explanations.
- Webinars are pretty popular in Centers of Excellence. COOP helped to organize in the CoE EXCELLERAT a the Modern Fortran webinar on May 4, 2021 .
Corrections and updates
We will keep this page updated according to our progress. More to come on HPC and ML if we have readers and time to do it :). Feel free to contact us for more informations.