ABSTRACT

Anders Logg Simula Research Laboratory and Department of Informatics, University of Oslo

Kent-Andre Mardal Simula Research Laboratory and Department of Informatics, University of Oslo

Martin Sandve Alnæs Simula Research Laboratory and Department of Informatics, University of Oslo

Hans Petter Langtangen Simula Research Laboratory and Department of Informatics, University of Oslo

Ola Skavhaug Simula Research Laboratory and Department of Informatics, University of Oslo

Scientific applications that demand petascale computing power often involve very complex mathematical models. For continuum models based on partial differential equations (PDEs), the model complexity also tends to significantly amplify the associated software complexity. This is particularly the

A

case when the PDEs are solved by finite-element methods involving unstructured grids on parallel computers. The present chapter aims to present a new way of developing human-and machine-efficient finite-element software for challenging scientific applications.