Finite element (FE) analysis has become a vital tool in science and 
engineering. However, the efficiency of solving FE problems depends strongly on 
the properties of the underlying partial differential equation.
While approaches allowing to solve large-scale FE problems in modern high-
performance environments are well-established, the traditional approaches 
become inefficient or even break down entirely for strongly heterogeneous 
problems and anisotropies.
Recent developments, particularly the use of composite materials in aerospace 
applications, demands the efficient solution of such problems. This issue 
becomes even more severe when applying uncertainty quantification methods.
The topic of this talk is the GenEO spectral coarse space which can be used to 
either construct preconditioners that perform well in these situations or to 
generate efficient coarse representations in multiscale applications.
A highly scalable implementation in the DUNE numerical framework will be 
presented as well as numerical results demonstrating its effectiveness.