Différences
Ci-dessous, les différences entre deux révisions de la page.
| Les deux révisions précédentes Révision précédente Prochaine révision | Révision précédente | ||
|
en:developpement:productions:logiciels:incns [2013/05/17 06:27] cicaluga |
en:developpement:productions:logiciels:incns [2015/01/07 10:04] (Version actuelle) |
||
|---|---|---|---|
| Ligne 1: | Ligne 1: | ||
| ====== Logiciel INCNS - Parallel INCompressible Navier-Stokes solver ====== | ====== Logiciel INCNS - Parallel INCompressible Navier-Stokes solver ====== | ||
| - | The code **INCNS** solves the incompressible Navier-Stokes equations in a cubic box with periodic boundary conditions in the three directions. It treats both Eulerian and Lagrangian dynamics. An external force ensures an isotropic stiring at low wavenumbers. The code is primarily devoted to the study of homogeneous and isotropic turbulent dynamics. | + | |
| + | * **Contact** : Emmanuel Lévêque, Cerasela Calugaru \\ | ||
| + | * **Objectif** : Portage du code sur GPU avec CUDA \\ | ||
| + | |||
| + | The code **INCNS** solves the incompressible Navier-Stokes equations in a cubic box with periodic boundary conditions in the three directions. It treats both Eulerian and Lagrangian dynamics. An external force ensures an isotropic stiring at low wavenumbers. The code is primarily devoted to the study of homogeneous and isotropic turbulent dynamics. | ||
| + | |||
| The integration is second-order in time (Adams-Bashforth scheme) and pseudo-spectral in space. The integration of fluid trajectories relies on a second-order (in time) Verlet algorithm and tri-cubic interpolation | The integration is second-order in time (Adams-Bashforth scheme) and pseudo-spectral in space. The integration of fluid trajectories relies on a second-order (in time) Verlet algorithm and tri-cubic interpolation | ||
| in space. | in space. | ||
| + | |||
| The code runs in parallel (with MPI) in either single or double precision with optional dealiasing. | The code runs in parallel (with MPI) in either single or double precision with optional dealiasing. | ||
| + | |||
| + | **Developments at CBP :** Implementing the Fast Fourier Transform by using CuFFT library on GPU (in PGI CUDA Fortran Compiler environment) instead of on FFTW on CPU. Tests of performance on NVIDIA cards (Tesla M2070, ...) | ||