Example PhD

Parallel Programming Paradigms for Modern HPC Platforms

Supervisor: Professor D.W. Walker

Keywords: MPI, OpenMP, computational chemistry.

In the current rapidly changing landscape of high performance computing (HPC) it is not clear which software paradigms are best able to deliver high parallel efficiency and scalability. Computational chemistry codes, such as NWChem and Molpro, often use the Global Arrays (GA) software as an abstraction for supporting a certain form of shared memory, but preliminary work has shown this may have problems in scaling to very large parallel systems for which other approaches, such as MPI and OpenMP (or a combination of both), may yield better performance. Support for GA will be removed from selected parts of the NWChem code and replaced with MPI and OpenMP constructs. Comparative performance studies will then investigate the scalability of these codes on HPC resources at HPC Wales and elsewhere (as available). It is also intended to extend existing work to perform core computational kernels of the NWChem code, such as BLAS 3 operations, on general-purpose GPUs.

Key Skills/Background: Open to Computer Science and computational chemistry graduates and postgraduates with good programming skills.

Contact: Professor D.W. Walker to discuss this research topic.