ACCU London

Thursday 17th July 2008, 19:00

The Challenges facing Libraries and Imperative Languages from Massively Parallel Architectures [Slides]

The challenges of writing programs for parallel architectures have become ever more prominent in the light of the pervasiveness of desktop multi-core processors and readily available consumer grids. This may be seen as an effective increase in instruction retirement rate via the multiple cores of execution that have been made available to the programmer. This has, in turn, exacerbated the memory wall, the rate at which data can be read from and written to main memory via the various memory hierarchies of super-scalar processors.

These massively parallel architectures may have sophisticated memory models, for which there is an open question regarding what may be the ideal approach to implementing parallelism, via using many threads, from the programmer's perspective. The issues raised by these massively parallel architectures are relevant to current systems: blade frames may contain between 4-8 processors per blade and have between 5-20 of these blades per frame, giving of the order of 20-160 processors. This is a reasonably large quantity of parallelism, in a NUMA architecture, for the programmer to manage.

Colin's and Jason's talk will introduce the problems associated with massively parallel architectures, and suggest techniques for developing software to cope!

About the Speakers

Jason M. McGuiness

Jason McGuiness graduated from the University of Bristol with a joint BSc (Hons) in Mathematics and Physics in 1990. Since graduating from Bristol, Jason has worked for a number of well known international companies as a technical architectural/software developer. Jason is currently working as a Manager at Barclays Capital where he is a Front Office Senior Developer.

In 2005, Jason took a year's break from the rigours of professional work and undertook a period of research on 'The Challenges of Writing Software for Massively Parallel Architectures. Jason was awarded an MSc(Research) from the University of Hertfordshire in 2006 for this work.

Colin Egan

Colin Egan graduated from the University of Hertfordshire, with a BSc (Hons) in Computer Science (Systems Engineering) in 1996. Subsequently to this, he was awarded a PhD in Computer Architecture in 2000, also from the University of Hertfordshire.

Previously, Colin worked for the National Health Service in the U.K. specialising in Clinical Microbiology (1978 - 1991). In 1991, he joined the University of Hertfordshire as a research assistant in the Department of Natural Sciences, researching into 'Neurotoxicology alternatives to animal testing.'

Since completion of his BSc in Computer Science he made a career change and has been working on high performance processing, multiprocessor systems and energy conservation computing. Colin has gained an international reputation for his work and is a member of a number of international program committees. He has over twenty-five published papers, mainly in Computer Architecture, some in Neurotoxicology and some in teaching and learning.