ZURUECK HOCH VOR INHALT SUCHEN

» Back to overview
Proposing Institution

Max Planck Institute for Astrophysics
Project Manager

PD Dr. Ewald Müller
Karl-Schwarzschild-Str. 1
85748 Garching
Abstract
GRMHD numerical codes have been extensively used in the last decade to investigate the properties and phenomenology of accretion flows onto compact objects (such as neutron stars and black holes). A few examples of publicly available codes are Athena, Harm and Whisky.With our code ECHO we are currently studying the stability of thick magnetized accretion disks orbiting around black holes, and in particular we want to focus on the development of global non-axisymmetric instabilities and how they are affected by the inclusion of the non-ideal effect given by turbulent resistivity: this feature distinguishes ECHO from the above mentioned codes, and we believe it will significantly enhance our knowledge of accretion disks. Since 3D GRMHD simulations are computationally quite expensive (even more when non-ideal effects are included) a good parallelization scheme and a general optimization of the code performance are crucial factors in the continuation and extension of such studies.A first improvement we expect is a significant increase of the time-stepping efficiency in our code. Most of the time is currently spent in doing a high number of floating point operations, which could probably be better optimized.We would also like to couple our current multidimensional MPI domain-decomposition scheme to an OpenMP parallelization applied to those parts of the code that are most suitable for it. Since we aim for a good behavior at large scaling, it will also be important to verify possible bottlenecks that may rise in this regime. This will then require an accurate check on the efficiency of the I/O treatment, communication strategies and in general a deep profiling of the code, so to pin down the parts where it spends most of the time and improve them as much as possible.

Impressum, Conny Wendler