Modern Domain Decomposition Methods in Both Space and Time
Master thesis, HiWi Position, Projectseminar, Bachelor thesis
Parallel algorithms play an increasingly vital role in both research and industry for accelerating simulations. Domain Decomposition methods (DDMs), such as nonoverlapping Schwarz or Dirichlet-Neumann/Neumann-Neumann methods, are effective approaches for introducing spatial concurrency, thereby facilitating parallel computations. When the problems under examination are not only spatially dependent but also time-dependent, waveform relaxation enhances information exchange between different time steps [2, 3]. This, in turn, enhances additional concurrency in time when combined with Parareal methods.
The objective of this project is to implement selected algorithms from the provided references and evaluate their performance using discretized benchmark problems.