Project Summary
|
|
Summer Scholarship Programme Project Summary |
|
|
|
Rob Baxter, EPCC
The physical problem is the calculation of the flow of geomagnetic fluids at the Core-Mantle Boundary based on measurements of variations in the Earth's magnetic field at the surface. Various models exist to do this, one of which is implemented in the ITERTRAN code.
A key feature of these calculations is the use of spherical harmonic expansions for the solutions to the differential equations (differential equations on a sphere yield spherical harmonic solutions). Computations are performed in terms of degrees and orders of associated Legendre polynomials rather than spatially-distributed field solutions.
Part of the previous work was the introduction of a spherical transform method into the existing ITERTRAN code, to perform intermediate transformations from the spherical harmonic space to a spatial grid upon which the solution can be integrated numerically.
Two options for parallelising the ITERTRAN code were considered: a domain decomposition of this intermediate spatial grid, or a task-farmed approach over independent timeslices at a higher level of the code.
It is proposed that this project investigate the implementation of one or both of these parallel models, either separately or in combination. Initially it is expected that the task-farmed approach over timeslices will prove most effective, although the parallelisation benefits for the spatial grid will rise for problems involving higher spherical harmonic truncation levels. It should prove interesting to investigate the interaction of these two complementary approaches to parallelisation.
The primary goal would be a task-farmed implementation based on MPI.
Additional goals would be an independent domain-decomposition version and a version that coupled the two approaches.
A possible further extension is the investigation of the use of a data-parallel approach to the domain decomposition version.
|
Webpage maintained by
mario@epcc.ed.ac.uk
|
|