|Position Title||Geophysics CUDA Developer|
|Summary||The undergraduate student will work with the MPI-based geophysical simulation code CitCOMs and work to port it to a CUDA-based hybrid platform. The student will test and benchmark his work using TeraGrid resources.|
|Job Description||Only a handful of computational geophyiscal studies consider the profound effects of a mineral phase change, such as those that occur in minerals at 670 km depth, a boundary that divides the Upper and Lower Mantle in Earthâs interior. Current computational models, with spatial resolutions around 100 km, largely ignore the presence of a phase change. To accurately model the behavior of minerals at this boundary and the effect it has on descending surface slabs, we require a much finer spatial resolution of 1-10 km, demanding significantly more computational throughput.|
If the program can be reorganized in a way that simple individual calculations can be done across large data sets, then the program could see significant improvements in throughput by executing it on a graphics processor using a technology such as nVidia's CUDA.
In scientific computing, increasing the resolution (thereby increasing the problem size) can often be particularly well suited to this form of refactoring. Rewriting these programs to use the capabilities of GPUs has the capacity to greatly increase the number of calculations that can be done, which will enable geologists to get the higher resolution they need to get accurate results.
|Location||University of Wisconsin - Eau Claire Department of Computer Science|
located in Eau Claire, WI (about 80 miles East of Minneapolis/St. Paul)