sabide.github.io
Stéphane Abide bio photo

Stéphane Abide

Associate Professor

Perpignan, France ResearchGate Github e-Mail

Research interest


High-Perpformance Computing for Higher-Order discretisations

The compact finite difference formulation is a numerical discretisation used for computing finite difference approximations. This method is known for the small computation stencil while allowing accurate approximations. The compact schemes has the advantage of favorable dispersive error and dissipative error properties when compared to explicit schemes. However, this method has a drawback as it is implicit and requires solving a diagonal matrix system to evaluate interpolations or derivatives at all grid points. So it remains a challenge to solve elliptic problems optimally with modern HPC… part of my work is devoted to this. page under construction

HOCS-cube a research code

For 5 years I have been developing a computational code dedicated to the high-performance numerical simulation of incompressible flows in shoeboxes ( referring to the computational domain). With this specific tool for compact schemes one can solve the incompressible Naviers-Stokes equations in an *old school way*: semi implicit time scheme, full-staggered. This allows me to test new ideas about compact schemes, especially the fact that they can be HPC-compliant rPDD . [JCAM] [CaF],

SRI - Baroclinic waves

High performance computing is motivated by the challenges of fluid physics. In this project led by Uwe Harlander, we try to perform both experiments and numerical simulations to demonstrate that the baroclinic tank can be a useful model for atmospheric flows. This flow is known to be difficult to simulate due to its multi-scale characteristics (IGW, 2d turbulence). In addition, heat transfers at the free surface can modify the dynamics of the baroclinic wave, making it more difficult to compare numerical calculations with experiments. [EGU] [GAFD]

Chebyshev-collocation spectral method

Over the last few months I have been working on the implementation of this method in HOCS^3. I am focusing on the differentially heated turbulent cavity at high Rayleigh number to assess LES SVV model. The first results are encouraging.