For few decades, numerical simulation methods have known several significant breaks ( trend towards massively parallel computing , transition to non-structured meshes, generalization methods of multi-physics coupling , ... ) and new changes about automatic adaptive mesh and multiphysics monolithic approximations are coming.
Models are multi-scale in space and time and solving systems are becoming stiffer. The user will be able to specify the number of unknowns used but has to take into account the poor packaging systems to control the cost calculation ( on parallel machine ). INRIA teams and Lemma are experts in parallel algorithms and adaptive mesh and want to achieve the junction between parallel multigrid and adaptive anisotropic mesh.
The theoretical part will synthesize the new adaptive formulations based on the norm, entered by the crew, and the theory of Full-Multigrid algorithm convergence. An important methodological research will be conducted on massively parallel scalable multigrid formulations. Scalability is the ability to compute 10 times faster on 10 times more processors. Digital applications will concern multiphase computations and fluid-structure terms from the platform ANANAS of Lemma Company.