The goal of this event is to bring together scientists from both the academic and industrial world around a challenging topic in scientific computing. During the first week, a classical summer school is proposed. It consists of several lectures given by leading scientists. The remaining five weeks are dedicated to research projects, after a morning seminar. During the research session, each participant works within a team on a project proposed by either an industrial or an academic partner. The teams typically consist of young researchers supervised by senior researchers. Experience from previous years has shown the significant positive impact of CEMRACS not only on the development of these projects, but also on the interaction between academia and industry.

The objective of CEMRACS 2016 is to better understand the new methodological challenges imposed to scientific computing by the evolution of the computer architecture. More precisely, we have identified 3 critical and complementary research fields that have to be re-thought to benefit from current and emerging parallel computers:

• Solvers and communication-avoiding algorithms in high performance computing
• Numerical methods suitable for parallel architecture: domain decomposition, parallelism in time.
• Data assimilation strategies compatible with parallel architecture