Contribution to a conference proceedings/Contribution to a book

FZJ-2018-02922

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Scalable Algorithms for Adaptive Mesh Refinement: Extension to General Element Types and Application to Fluid Dynamics

Burstedde, C. ; Calhoun, D. ; Fonseca, J. A. ; Holke, J. ; Kollet, S.FZJ* ; Lahnert, M. ; Mehl, M.

2018
Forschungszentrum Jülich GmbH, Zentralbibliothek Jülich

Please use a persistent id in citations: http://hdl.handle.net/2128/18504

Abstract: We discuss the development of parallel algorithms for adaptive mesh refinement (AMR) and their application to large-scale problems in simulating fluid dynamics. Our approach to AMR can be described as using a forest of octrees (or quadtrees in 2D) that are adaptively refined. The storage of elements is distributed in parallel, and fast and scalable algorithms exist for dynamic refinement/coarsening and other important tasks, such as partitioning and the extraction of one layer of off-process (ghost) neighbours. Our contributions are twofold: (a) We use the p4est software as a basis to create numerical applications to simulate the flow of gas in the atmosphere (advection equations), variably saturated subsurface flow (Richards), and the free flow of liquid (Navier-Stokes). (b) In addition to using quadrilateral/cubic elements, we are developing space filling curves and high-level AMR algorithms for triangles and tetrahedra. We include scalability results and simulation snapshots obtained on the JUQUEEN supercomputer.

---

Contributing Institute(s):

1. John von Neumann - Institut für Computing (NIC)
2. Agrosphäre (IBG-3)

Research Program(s):

1. 255 - Terrestrial Systems: From Observation to Prediction (POF3-255) (POF3-255)

Appears in the scientific report 2018

---

Database coverage:

---