Development and implementation of parallel adaptive algorithms
for computational simulation of electromagnetic scattering
The successful candidate will have an undergraduate or postgraduate training
in one of the following areas: mathematics, computer science, engineering or
similar numerate discipline and will be keen to participate in development and
implementation of novel numerical algorithms for computational simulation using
state of the art high performance computers. As such, you should ideally have
experience of programming in a high level language such as C, C++, FORTRAN or
similar. Experience using parallel computers would be an advantage but is not
required since training can be provided. The project, supervised by
Prof. Mark Ainsworth and Dr Ozgur Ergul, will be based in the Department of
Mathematics and Statistics, and will focus around the use of finite and boundary
element methods for the simulation of electromagnetic scattering.
The project is funded in part by Strathclyde University, and by the Centre
for Numerical Algorithms and Intelligent Software (www.nais.org.uk) a recently
formed large multidisciplinary research centre based at the Universities of
Strathclyde, Edinburgh and Heriot-Watt. This exciting new centre provides a
unique opportunity through which the student will have access to advanced
training and state of the art hardware.
For further information please contact either
Prof. Mark Ainsworth (M.Ainsworth@strath.ac.uk)
or Dr Ozgur Ergul (ozgur.ergul@strath.ac.uk).
Online application forms can be found at
http://www.maths.strath.ac.uk/applying/postgraduate/applications
in which you should quote Ref. Ainsworth/Ergul NAIS Studentship.
link
This PhD project involves the development and analysis of approximate methods for partial and stochastic differential equations, along with the linear algebra needed to make them work efficiently. The goal is to develop highly efficient numerical algorithms suitable for modern high performance computers, from the now commonplace desktop dual and quad-core computers, through the upcoming multi-core desktop machines and on to high-end national research laboratory scale computers. The equations arise in the modelling of real-world problems (e.g. in the energy sector, medicine and biology) and the project will involve interactions with the end-users.The studentship is provided by Heriot-Watt University in support of the NAIS project.
We offer a lively, supportive and highly research-active environment for graduate students in Mathematics at Heriot-Watt, embedded in the Maxwell Institute for Mathematical Sciences partnership with the University of Edinburgh as well as the NAIS project. Students with a good first degree or MSc in mathematics with knowledge of numerical analysis, differential equations and some analysis would be ideal. Training is available in numerical analysis and in high performance computing. link
A grant was awarded following the EPSRC HPC Algorithms Sandpit to a team from the Universities of Manchester, Leeds, Edinburgh, Hull and Strathclyde. Dr Mark Bull from NAIS partner EPCC will lead the Edinburgh part of the project. Summary: The project aims to enable a step change in the performance of a wide range of real world applications, by applying and improving existing asynchronous algorithms for large sparse systems, and developing and analyzing entirely new asynchronous algorithms. The project will develop prototype software implementations of these algorithms which demonstrate both scalability and reliability. In the implementation of the software we will address the diversity of hardware platforms and programming models which will be found in future generations of machines, with at least 1 million cores. Another key aspect will be to exploit the inherent fault-tolerant nature of asynchronous algorithms to help address the need to cope with frequent hardware failures in extremely large machines. The software will be integrated into our two exemplar applications (bone modelling and power grid management), enabling them to carry out entirely new science as a result of enhanced problem resolution and reduced time to solution. link
NAIS will provide HPC training for all staff connected to the project, and, on a space available basis, staff from other institutions. Dates: 15 February, 22 February, 1 March 2010. There will also be a more in-depth training workshop in Summer 2010. See the links on the NAIS Events page link
Andreas Grothey (Edinburgh-Mathematics) will act as deputy to Iain Duff (Rutherford Appleton Lab) on the European Exascale Software Initiative, organised by a consortium of European leaders in HPC. The main goal of EESI is to build a European vision and a roadmap to address the challenge of Peta and Exascaling computation and simulation. This will be done by involving large numbers of European actors in the HPC sphere, investigating how Europe is located in the overall international HPC landscape and competition, its strengths and weaknesses, and the topics and challenges that Europe may and must develop. EESI will also identify the sources of competitiveness for Europe induced by the development of Peta/Exascale solutions and usage.
link
The Edinburgh Parallel Computing Centre, on behalf of NAIS, is running a survey of UK HPC Numerical Algorithms in order to better target its resource and projects to themes and projects of UK significance. This survey is being conducted by Lawrence Mitchell and George Becket at EPCC. In coming weeks they will be contacting many users and developers of HPC codes and numerical analysts, both at the NAIS constituent universities, NAIS network partners, and beyond, to guage their needs and interests.
NAIS has received matching funds from the UK's e-Science Institute and the EPSRC-funded Research Network on Mathematical Challenges of Molecular Dynamics (MD-NET) to run an international meeting June 30-July 3, 2010. From the overview of the meeting: "...As uniprocessor performance increases tail off, the search is on for new types of (molecular modelling) methods that can continue to offer speed increases in the setting of distributed hybrid architectures, or the grid. An interesting recent development which is becoming prominent in the molecular simulation literature is the reformulation of molecular modelling in terms of path sampling strategies, whereby multiple trajectories are used to probe complex configurational pathways, often disturbing thermodynamic equilibrium to enhance mobility or to force rare transitions. These multiscale molecular algorithms are well suited to implementation in an event-driven distributed computing setting... "
link
Magnus Svard, previously based in Oslo, has been hired to a NAIS Lectureship at the University of Edinburgh. Magnus works on numerical analysis problems related to computational fluid dynamics. Ozgur Ergul from Bilkent University (Turkey) has been hired to a NAIS Lectureship at the University fo Strathclyde. He works in parallel simulation of electromagnetic problems. Vijay Nagarajan has taken up a NAIS Lectureship in the School of Informatics at the University of Edinburgh. Vijay studies compilers, computer architecture and software engineering, in particular in the setting of multicore technology.
link
NAIS's launch event took place on October 21. Coming on the heels of the NSF-NAIS Workshop on Intelligent Software: the Interface between Algorithms and Machines the Launch Event featured a superb public lecture by Marc Snir (University of Illinois) on Key Challenges in High Performance Computing: Algorithms and Programming Environments for Petascale and Beyond. This was followed by a fine reception in the new Informatics Forum and a dinner at St Leonards Hall. Photos Below from the Launch Event.
![]() |
| The stunning new Informatics Forum at the University of Edinburgh provided an excellent setting for the Launch Event. |
![]() |
| Marc Snir demonstrates Illinois-style parallellism. |
![]() |
| Cora Cartis (Edinburgh-Maths) discusses with Phillippe Toint (NAIS Advisory Board) and Nick Gould (Rutherford Appleton Lab). |
![]() |
| Liz Jessup (Colorado) and Iain Duff (Rutherford Appleton Lab). |
![]() |
| Photo from NSF-NAIS Workshop on Intelligent Software |
![]() |
| At the reception |
![]() |
| Strathclyde's Dean of Science Iain Hunter and Edinburgh's Head of College of Science and Engineering Nigel Brown (PI on the NAIS project). |
![]() |
| Ben Leimkuhler (NAIS Director) in discussion with Rudolf Roemer, the director of Warwick's Centre for Scientific Computing. |