A Modular Design for Parallel Adaptive Finite Element Computational Kernels

The paper presents modular design principles and an implementation for computational kernels of parallel adaptive finite element codes. The main idea is to consider separately sequential modules and to add several specific modules for parallel execution. The paper describes main features of the proposed architecture and some technical details of implementation. Advanced capabilities of finite element codes, like higher order and discontinuous discretizations, multi-level solvers and dynamic parallel adaptivity, are taken into account. A prototype code implementing described ideas is also presented.