This page contains a Flash digital edition of a book.
In 2008, Jaguar, a Cray XT5 system located at the Oak Ridge Leadership Computing

Facility, became the world’s fi rst supercomputer to break the petafl op performance barrier. Since then, the high-performance computing (HPC) community has been

working towards the exafl op machine – a project supported by the US Department of Energy and the US National Nuclear Security Administration. Both agencies have set a timeframe of 2018 for the development of exascale systems, capable

of running more than 1,000 times faster than any of today’s machines. Industry experts give their opinions on whether, and how, we might reach this goal

Pete Beckman, co-leader of the International Exascale Software Project and director of Argonne’s Exascale Computing and Technology Institute

laptops or desk-side machines might have had one CPU, or even two, but now chip technologies are shifting to multicore. This fundamental change means that programming and supercomputer architectures must adapt as we look to the next generation. Computer scientists tend to think in multiples of 1,000, so we go from Pfl ops (1015 (1018

‘O ) to thinking about exascale

) and we realise the issue of electrical power and the shift to multiple cores will


ne of the most important things happening at the moment is the change in architecture for processors. In the past our

dramatically change the architecture and programming of these systems. ‘Programming is a challenge, and in

some ways we look at supercomputers as a forward-looking time machine that represents the technology that will eventually be commonplace in desktop systems. Currently, to use the largest supercomputer, roughly 100,000 different pieces of a parallel program must all running simultaneously. That means to create a simulation of combustion within a jet engine, for example, the computation would need to be divided into 100,000 pieces that could all be run at the same instant on different CPU cores.

That’s quite a challenge. However, as we look to exascale we realise that we will need to move to millions or maybe even billions of threads of control. The way we write programs and develop software is typically slow to change, but it will have to, and soon, as this change is upon us right now. It’s causing a little bit of concern in the community as we understand the complexity and magnitude of this paradigm shift.’

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44