HPC Short Courses - a coordinated response for training the UK's next generation of HPC users

This is a proposal for a programme of short courses designed to provide an advanced training programme for High Performance Computing (HPC) that is both rigorous and wide ranging. The programme is designed as a partnership with PhD supervisors, freeing the supervisor to focus on the more discipline-specific aspects of their student's training. We will provide three types of activity: (1) an annual 2-week residential course (Autumn Academy) designed to allow students to start working with HPC code right from the start of their PhD; (2) A rolling 2-year programme of short residential courses on more specialised topics and that builds from the base provided by the Autumn Academy; and (3) a programme of half- or one-day events, using Access Grid technology where possible, designed to broaden the student's background in HPC related issues while also helping them to form networks with their peers. The proposed training programme is built firmly on the very considerable experience of the applicants and their 13 institutions in HPC training.

The ability of high performance computing (HPC) to model complex systems with increasingly high fidelity means that its impact is growing very rapidly, not only within academia but also within government and industry. Traditionally, the areas of application of HPC were Science and Engineering, but that is no longer the case; HPC is increasingly finding applications in the Social Sciences and Economics, and the Life Sciences and Medicine. Thus the potential beneficiaries of developments in software for HPC are very wide-ranging. The benefits of this, however, can not be attained unless there is a highly skilled cohort of people able to write the code for these applications, and more importantly, to continue to adapt the algorithms and code to exploit the coming generations of computer architecture. It is this human infrastructure that will be developed by our training programme, and thus will ensure that these wide-ranging benefits are fully realised within the UK. Further, the architectural developments of HPC systems today will undoubtedly shape the architectures of personal computing systems of the near future; increasing parallelism (and heterogeneity) will pervade both high-end, and mainstream computing platforms. The developments that result from this project will impact both HPC users of today, and mainstream computing users of tomorrow. This project will have considerable impact on all these areas, as the ability to exploit HPC in Scientific Computing, the Social Sciences, Economics and the ability to shape future personal computer systems all depend crucially having highly skilled programmers who are familiar with the opportunities and constraints of current architectures, and who are looking ahead to future developments in computer architecture. These are precisely the skills that this project will help develop, with consequent benefits to both UK industries and companies, and academia. We will achieve this impact by training the most promising students, and by defining the nature of HPC training that is needed for the coming decade. This impact will be distributed not just through the students we train (and their supervisors), but also through the links we establish with UK industry.