Fast scalable solvers for robust discretizations in CFD
Rüde U (2018)
Publication Language: English
Publication Type: Conference contribution, Conference Contribution
Publication year: 2018
Event location: Glasgow
Abstract
The talk discusses exa scale computing topics for CFD, including issues such as energy consumption, algorithmic complexity, and scalability. Additionally, a novel approach to the design and analysis of discretization methods was recently introduced. The so-called Hybrid High Order (HHO) discretization relies on degrees of freedom that are broken polynomials on a general mesh and its skeleton. This advanced method promises several attractive features including better conservation of physical properties, and has been successfully applied to the discretization of several linear and nonlinear problems. However, research on HHO methods has primarily concentrated on their approximation properties and the question of fast, scalable linear solvers has not been treated yet in the literature. In contrast to lower order discretizations, the system matrices are no longer diagonally dominant, making classical Jacobi or Gauss-Seidel methods inefficient as smoothers. We present research on scalable multigrid techniques and discuss their potential for these novel discretization.
Authors with CRIS profile
How to cite
APA:
Rüde, U. (2018). Fast scalable solvers for robust discretizations in CFD. In Proceedings of the ECCM-ECFD 2018: Minisymposium. Glasgow, GB.
MLA:
Rüde, Ulrich. "Fast scalable solvers for robust discretizations in CFD." Proceedings of the ECCM-ECFD 2018: Minisymposium, Glasgow 2018.
BibTeX: Download