HR: 0800h
AN: SM51B-0374    [Abstracts]
TI: Parallel, time asynchronous modeling of plasmas: Overcoming the computational challenges in traditional (MHD, Hall MHD, Full Particle, Hybrid, Vlasov) codes
AU: Omelchenko, Y
EM: yurio@scibernet.com
AF: UCSD, 9500 Gilman Drive, Dept. of ECE, La Jolla, ca 92093-0407 United States
AU: * Karimabadi, H
EM: homak@scibernet.com
AF: UCSD, 9500 Gilman Drive, Dept. of ECE, La Jolla, ca 92093-0407 United States
AU: Driscoll, J
EM: jdriscoll@copper.net
AF: UCSD, 9500 Gilman Drive, Dept. of ECE, La Jolla, ca 92093-0407 United States
AU: Dave, J
EM: jagrut@cc.gatech.edu
AF: Georgia Institute of Technology, College of Computing, Georgia Tech, Atlanta, GA 30332 United States
AU: Fujimoto, R
EM: fujimoto@cc.gatech.edu
AF: Georgia Institute of Technology, College of Computing, Georgia Tech, Atlanta, GA 30332 United States
AU: Perumalla, K
EM: kalyan@cc.gatech.edu
AF: Georgia Institute of Technology, College of Computing, Georgia Tech, Atlanta, GA 30332 United States
AU: Omidi, N
EM: omidi@scibernet.com
AF: UCSD, 9500 Gilman Drive, Dept. of ECE, La Jolla, ca 92093-0407 United States
AB: Computer simulation of many important complex physical systems such as the Earth's magnetosphere has reached a plateau because most conventional techniques are ill equipped to deal with the multi-scale nature of such systems. The traditional approach to modeling spatially distributed physical systems has been based on time-driven (or time-stepped) simulations (TDS) where the whole state of the system is updated synchronously at discrete time intervals. This method has two inherent inefficiencies with severe consequences: (i) the well-known time step restriction imposed by a global CFL (Courant-Friedrichs-Levy) condition, (ii) uniform (and unnecessary system update) computational work independent of the level of activity in a given region. We have been working on an entirely different (asynchronous) simulation methodology based on a discrete event-driven (as opposed to time-driven) approach. Our ultimate goal is to develop a 3D global multi-physics code for application to Earth's magnetosphere. Here we report on our progress where we have developed a general parallel infrastructure based on this new technique. We demonstrate the power of this technique through a 1D parallel hybrid simulation of a fast magnetosonic shock. We find that the code is over a factor of 30 faster than the traditional hybrid codes. In our technique, individual parts of the global simulation state are updated on a "need-to-be-done-only" basis and all simulation entities (individual particles/phase space elements/fluid elements, local fields) evolve on their own physically determined time scales. This has immediate implications for all types of plasma simulations. For example, one of the obstacles to the use of Vlasov codes in 2D and 3D is the fact that most of phase space is inactive but still has to be carried in the computation using standard techniques. This inefficiency makes the Vlasov codes almost unusable in 3D where the phase space (consisting of three spatial coordinates and three components of velocity) is very large. In our technique, only the "active" regions of phase space are updated. Another example is in regards to simulation of interplanetary shocks and physics of particle acceleration which remains beyond the scope of existing hybrid codes due to the multi-time scale nature of the problem.
DE: 7851 Shock waves
DE: 2100 INTERPLANETARY PHYSICS
DE: 2700 MAGNETOSPHERIC PHYSICS
DE: 2753 Numerical modeling
SC: SPA-Magnetospheric Physics [SM]
MN: 2004 AGU Fall Meeting