Performance Portability in Scientific Applications - developing parallel programming models for real scientific applications which can address the issue of "performance portability". at University of Edinburgh on FindAPhD.com

This project is no longer listed on FindAPhD.com and may not be available.

Click here to search FindAPhD.com for PhD studentship opportunities

Dr K Stratford No more applications being accepted Competition Funded PhD Project (UK Students Only)

About the Project

Many scientific applications are extremely complex and have very long development times. They are almost all written in "traditional" programming languages such as C/C++/Fortran. This has meant that few have adapted gracefully (or at all) to increasingly heterogeneous high performance computing infra-structures incorporating highly vectorised many-core architectures, graphical processing units, and occasionally more exotic hardware such as digital signal processors and field programmable gate arrays. The challenge is therefore the development of programming models which allow scientific application developers to write portable and efficient applications in an expressive manner and without excruciating pain!

A PhD in this area might address the following questions:

* Many scientific applications achieve a disappointingly low performance compared with the peak performance hardware can deliver. This is a waste of energy and money. Can improved programming models, compilers, and tools help to solve this "performance deficit"?

* Can such performance be achieved in a portable manner on different types of hardware? What types of hardware are most cost-effective?

* Can programming models express a move away from a "bulk synchronous" execution to something more directed toward the asynchronous execution of independent tasks?

The aim is help to develop parallel programming models for real scientific applications which can address the issue of "performance portability". This might include development of the model itself, development of new algorithms, tools to aid analysis of performance portability, and implementation of programming models on new hardware platforms.

Candidate requirements:

Applicants should have a strong programming background, and a degree in a numerate discipline. Experience in programming is essential, and prior exposure to high performance computing would be beneficial. Training will be provided.

About The Centre:

As a PhD student you would be based at Edinburgh Parallel Computing Centre (EPCC), part of the University of Edinburgh, where we have close ties with both The School of Informatics and The School of Physics and Astronomy. EPCC runs a number of national supercomputing services, and has strong links with manufacturers such as Cray, Intel, and NVIDIA.

Funding Notes

To be eligible for funding covering both fees and stipend, students must be UK nationals, or EU nationals who have been resident in the UK for at least 3 years before commencing the studentship.

References

[1] For an overview of technical issues, see, e.g., the United States DOE Centers of Excellence Performance Portability Meeting Final Report (2016): https://asc.llnl.gov/DOE-COE-Mtg-2016/talks/COE-PP-Meeting-2016-FinalReport.pdf

[2] A Lightweight Approach to Performance Portability with targetDP, A. Gray and K. Stratford. International Journal of High Performance Computing Applications 1-14 (2016).

Where will I study?

University of Edinburgh

One of the world''s top universities, consistently ranked in the world top 50 and placed 20th in the QS World University Rankings 2020. We provide a stimulating working, learning and teaching environment with access to excellent facilities and attract the world''s best, from Nobel Prize winning laureates to future explorers, pioneers and inventors.

Performance Portability in Scientific Applications - developing parallel programming models for real scientific applications which can address the issue of "performance portability".

University of Edinburgh Edinburgh Parallel Computing Centre (EPCC)