Numerical approximations of time-dependent partial differential equation (PDE) problems in three spatial dimensions often require very fine grid resolutions, and parallel computing can be employed to speed up the computations by subdividing the spatial domain over the available parallel processors. However, parallelization in space alone becomes inefficient on new generations of parallel computers where the number of parallel processors (or cores) is very large. In order to increase the concurrency and parallel efficiency, one can consider to carry out computations that iteratively improve the approximation at different time levels in a concurrent fashion. This approach is very attractive conceptually. In particular, multilevel methods for parallel computing in time will be explored for model problem PDEs of parabolic and hyperbolic type.

Relevant link:
-http://computation.llnl.gov/project/parallel-time-integration/pubs/mgritPaper-2013-3.pdf

Required:
-major in computational/applied mathematics, computer science, or engineering
-at least one course on numerical computing; a course on PDEs
-experience with programming (any of Matlab, Python, C, Java, C++, MPI, ...) and interest in parallel computing

Duration:
6 weeks, starting in December or January

Funding:
Subject to Funding by Monash University