Time-driven Simulations
|
|
|
|
We are working on the development of next generation kinetic plasma codes.
Central to our technology is the modularized, object-oriented API for automating
the development of massively parallel kinetic plasma simulations. The key
features of this technology are:
- A flexible and highly scalable parallel PIC framework that will enable
running very large (>100 million particles, 10 million grid cells, 100
thousand time steps) hybrid and full PIC simulations on 100s of distributed
memory processors.
- Built-in load balancing scheme that will address the most serious parallel
PIC code performance problem namely, processor load-imbalance due to dynamic
spatial and temporal inhomogeneities in particle algorithms combined with the
need to statically balance field update.
- Implementation of adaptive mesh refinement (AMR) on structured grids, and
efficient (Krylov, multi-grid) elliptic solvers for PIC
simulations.
A schematic of our modularized PIC architecture
(MPICA) is shown below:
MPICA will serve as the development infrastructure of our next generation
parallel hybrid code, called "PHYSICS" (Particle HYbrid Simulation via
Implicitly Coupled Scheme).
|
|
|
|
