 |
| Titel |
DMSwarm: Particles in PETSc |
| VerfasserIn |
Dave A. May, Matthew G. Knepley |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250146131
|
| Publikation (Nr.) |
 EGU/EGU2017-10133.pdf |
|
|
|
|
|
| Zusammenfassung |
| Whilst specific details of particle methods are as diverse as the applications in which they are utilized within, such methods have a number of common requirements. Specifically these entail: the need to attach data (or fields) to each point; support for the advection of points within some domain; dynamic insertion and deletion of points; collection type operations to gather nearby points; and methods to interpolate and restrict data back and forth between a set of points and a background mesh. Despite the commonality of these requirements, to date, few computational libraries provide high level support for particle methods in either a sequential, or parallel (MPI) computing environment.
In this presentation, we describe the recently introduced implementation DMSwarm within the Portable, Extensible Toolkit for Scientific computing (PETSc) which provides a fully parallel solution for pure particle methods (e.g. DEM, SPH, EFG) and for particle-mesh methods (e.g. PIC, FLIP, MPM, GIMP).
Particle-In-Cell (PIC) methods with geodynamic simulation tools are ubiquitous. To illustrate the functionality provided by DMSwarm, we present results from viscous flow calculations using a finite element PIC method applied to study Rayleigh-Taylor instabilities using both structured and unstructured meshes. Our results are derived from standard examples provided with the PETSc source distribution and serve as an entry-point for new users to learn how to incorporate DMSwarm into their geodynamic software. |
|
|
|
|
|
|