Astrophysical problems often consist of a complex interplay between physical processes. Most of the time, they involve a large range of scales, associated with rapid dynamics, that one wishes to study. Theoretical astrophysics therefore heavily relies on numerical models.
In the last few decades, various numerical codes have emerged to tackle these astrophysical problems. Usually, these codes are designed for parallelization on tens to millions of CPU cores. However, the current High-Performance Computing ecosystem is shifting towards new hybrid accelerated machines to reach both exascale performance and meet energy sobriety requirements. In order to take advantage of such accelerated systems, existing codes need to be rewritten, taking into account the specificities of the targeted architecture. To this date, no consensus has been reached on an architecture-independent programming standard. With the limited support and maintenance staff, our community cannot afford to develop one version of a code for each architecture.
The IDEFIX code was designed to address all these issues. It provides a user-friendly framework to run fluid simulations in parallel, both on CPU and GPU devices. It uses the metaprogramming library KOKKOS to remain highly performant while being portable.
Introduction to Development and Execution : a First Insight of idefiX
The code is now public and available on GitHub. The aim of these first IDEFIX user days is to provide hands-on tutorials for potential users and promote the code through its diverse scientific applications.
Practical information
The workshop will occur at IPAG from July, 4 to 6 2023. PhD students and postdocs are encouraged to apply. Limited financial support is available for PhD students and postdocs upon request to the organisers.
NB: registration and abstract submission will close on May 1st 2023.
Further information on the organization will be uploaded in the future.
Loading... 
