From hybrid architectures to hybrid solvers

International audienceSolving large sparse systems of linear equations is a crucial and time-consuming step, arising in many scientific and engineering applications. Consequently, many parallel techniques for sparse matrix solution have been studied, designed and implemented based on factorization or hybrid iterative-direct approaches. In this context, graph partitioning and nested dissection ideas have played a crucial role. The main goal of this presentation will be to give an overview of the continuum between these various algorithmic approaches and to present the improvements of the algorithms and of the associated parallel implementations in a manycore context. Numerical experiments on large irregular real-life problems will illustrate this work