Physics
Scientific paper
Dec 2009
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2009agufmsm51a1316o&link_type=abstract
American Geophysical Union, Fall Meeting 2009, abstract #SM51A-1316
Physics
[1956] Informatics / Numerical Algorithms, [2753] Magnetospheric Physics / Numerical Modeling, [2784] Magnetospheric Physics / Solar Wind/Magnetosphere Interactions, [2799] Magnetospheric Physics / General Or Miscellaneous
Scientific paper
Hybrid (electron fluid, kinetic ion) simulations have recently emerged in large-scale laboratory (spheromak, FRC) and global magnetospheric studies with the goal of accurately predicting energetic particle transport in complex plasma/magnetic configurations. Multiple temporal scales associated with both plasma and magnetic field inhomogeneities force severe restrictions on the global timestep in time-stepped hybrid simulations. We present an alternative computational tool for multi-scale plasma modeling: a first-ever uni-dimensional (2D or 3D chosen at compile time) asynchronous hybrid code, HYPERS (HYbrid-Particle Event-Resolved Simulation). HYPERS is currently being developed as part of the Virtual Hybrid Particle Laboratory (VHPL) in support of the SSX experiment at Swarthmore. It discards traditional time stepping in favor of the Discrete-Event Simulation (DES) approach [1]. Time increments for individual particles and local electromagnetic fields are adaptively selected through limiting per-update changes of their properties. This enables fast, reliable and accurate simulations of energetic plasmas immersed in highly inhomogeneous magnetic fields. Preliminary results from 2D simulations of the interaction of streaming plasmas with magnetic dipoles are discussed. We also report our undergoing efforts on developing efficient dynamic load-balancing strategies for future parallel HYPERS runs on petascale architectures. In this initial development phase, we are exploring tradeoffs of developing 2D/3D variations of extremely fast 1D load balancing heuristics such as chain-on-chain partitioning versus using fast geometry-based heuristics.
Brown Richard M.
Catalyurek U.
Karimabadi Homa
Omelchenko Y.
Saule Erik
No associations
LandOfFree
HYPERS: First Ever Multi-Dimensional Asynchronous Hybrid Simulations does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with HYPERS: First Ever Multi-Dimensional Asynchronous Hybrid Simulations, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and HYPERS: First Ever Multi-Dimensional Asynchronous Hybrid Simulations will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1880992