The evolution of dark matter particles is computed using a Particle-Mesh (PM) solver within an Adaptive Mesh Refinement (AMR) grid, while the Poisson equation is solved with a multi-grid method. Using a density based refinement criterion, the code dynamically zoom-in and follows all the overdense regions within the cosmological volume.
We have developed a new version RAMSES_DEUS of the cosmological code RAMSES (Teyssier, 2002) where the memory occupancy has been strongly improved (special thanks to P. Wautelet), some aspects of parallelization has been optimized and where we have included quintessence background evolution
The code is parallelized with MPI. It uses a dynamical Peano-Hilbert domain decomposition which allows a good scaling. We have used different configuration depending on the number of particles. The one billion particles simulations have run on 4096 Blue Gene/P cores at IDRIS supercomputing center for 5 000 000 mono-cpu hours. The 9 billions particles simulation have run on 24576 Blue Gene/P cores for 2 000 000 mono-cpu hours.
The main characteristics of the simulations are the following (Each simulation contains 6 refinement level and has been evolved down to z=0).
The main characteristics of the cosmological models are the following.