ANR LEGO / Ramses + Arpege Climat (Middleware)
Conducted byFrancois Broquedis, Jérôme Clet-Ortega
DescriptionRamses is a versatile platform to develop applications using Adaptive Mesh Refinement for computational astrophysics, developed by Romain Teyssier for the CRAL (Centre de Recherche Astronomique de Lyon). Arpege Climat is a high-resolution global model developed by Eric Maisonnave for the CERFACS (Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique). Both of these applications are implemented in Fortran 90, with some parts of Fortran 77 for Arpege Climat, and are part of the ANR LEGO project. As Master 1 students at Bordeaux 1 University, our job here is to set these applications up on the grid in order to run benchmark tests using MPICH 1.2.7, then using MPICH/MAD, a library which is part of the PM2 project developed by RUNTIME team (INRIA), and finally using PadicoTM, a communication framework for grids. Setting up multiple libraries is necessary to run these applications, and tests have to be done on different architectures and compilers. That's why we are using different sites of the grid, like Bordeaux, Toulouse and Rennes. The benchmark tests have to be launched on a large range of processors, at each step of the project evolution. At this point, several compiling problems have been resolved, but some of them still remain. Plus, some Fortran compilers used on the grid have license issues, like ifort on the Toulouse site.
Tools usedNo information
ResultsWe already successfully tested Ramses on up to 40 nodes, and we are still working on it.
More information here
Shared by: Francois Broquedis, Jérôme Clet-Ortega
Last update: 0000-00-00 00:00:00