Fast Solution of PDEs in Reservoir Simulation
using high performance computing to solve differential equations in
real-world problems? Want to learn more about achieving high
performance in parallel application codes? This project, working with
a major company in the oil industry, has two major parts. The first is to
develop faster implementations of the current methods for parallel
systems with 100-1000 nodes and 1000-10,000 cores, using a mix of MPI
and OpenMP (thread) programming. We are exploring a combination of
new data structures for sparse matrices and hybrid parallel
programming models. The second is to develop better
methods for solving the linear systems of equations that arise from
the PDEs used in reservoir simulation. Here we are exploring the use
of recent developments in multigrid methods.
We are looking for one or two graduate students interested in a
research assistantship to work on this project, starting immediately.
Contact: Bill Gropp
or Luke Olson.