Structured adaptive mesh refinement (SAMR) techniques aim
at preserving the high computational performance achievable on uniform grids
on a hierarchically adapted non-uniform mesh. The course will describe the
mathematical background of the typi-cally used discretizations, detail the
employed algorithms, and summarize practically relevant applications and their
implementation on modern parallel computers. Advanced topics such as using
the SAMR approach for implementing geometric multigrid algo-rithms for elliptic
problems and parallelization on massively parallel systems will also be covered.
A discussion of presently available SAMR software and a practical exercise
with the AMROC framework will complete the short course.
Outline:
- Fundamentals: Hyperbolic problems
- Finite volume schemes
- Discussion of mesh adaptation approaches
- Structured AMR for hyperbolic problems
- The recursive algorithm with time-space refinement
- Distributed memory parallelization
- Complex hyperbolic structured AMR applications
- Consideration of non-Cartesian geometries
- Shock-induced combustion
- Fluid-structure interaction
- Structured AMR for elliptic problems
- Adaptive geometric multigrid algorithms
- A parallel adaptive multigrid method for image registration
- Further topics
- Available implementations
- Exercise with the SAMR framework AMROC
This course is intended for summer students of Applied Mathematics
and Computational Sciences. If you want to participate, please send email
to deiterdingr@ornl.gov to allow
us a headcount.
Course material
Location: JICS, Bldg. 5100, Room 134
daily Monday 07/25/11 to Friday 07/29/11, 1:00pm-2:30pm
Friday 07/29/11, 3:00pm-4:30pm, practical exercise (Linux/Unix laptop required)