A million-core computer that programs itself could solve engineering problems hundreds of times faster than today’s supercomputers, at a fraction of the cost. Conventional machines still rely on a central processor doing one task at a time, even as they have grown to billions of transistors. This project, POETS, abandons that approach. Instead, the programmer defines a small set of simple behaviours for thousands of cheap processor cores, then lets the system “self-organise” to produce the result. A previous EPSRC project proved this works for neuroscience; POETS will test it on physics-based engineering problems. If successful, the technique could transform how we tackle compute-intensive tasks in fields such as structural engineering, fluid dynamics, or materials design. A simulation that now takes weeks on a traditional supercomputer might finish in hours on a POETS machine, using far less energy and hardware. The project also addresses a practical barrier: convincing industry to adopt a radically different computing paradigm, even when the performance gains are clear.
View original technical description
POETS (Partially Ordered Event Triggered Systems) is a significantly different way of approaching large, compute intensive problems. The evolution of traditional computer technology has taken us from simple machines with a handful of bytes of memory and (by the standards of today) glacial clock speeds, to multi-gigabyte architectures running five or six orders of magnitude faster, but with the same fundamental process at the heart: a central core doing one thing at a time. Over the past few years, architectures have appeared containing multiple cores, but exploiting these efficiently in the general case remains a 'holy grail' of computer science. POETS takes an alternative approach, made possible only today by the proliferation of cheap, small cores and massive reconfigurable platforms. A previous EPSRC project, BIMPA, enabled us to assemble a million core machine, creating a kind of 'meta-computer'. Rather than program explicitly the behaviour of each core and each communication between them, as is done in conventional supercomputers, here the programmer defines a set of relatively small, simple behaviours for the set of cores, and leaves them to get on with it - with the right behavioural definitions , the system 'self-organises' to produce the desired results. BIMPA was designed primarily for neuroscience applications, but a subsidiary research objective allowed us to study the use of the architecture for alternative (physics-based) problems, and we have demonstrated that this kind of approach can lead to dramatic speed increases over conventional solution techniques. POETS is not a general-purpose computing technique, but it is elegantly suited to a variety of traditionally compute intensive engineering and research problems, where it can produce results orders of magnitude faster than conventional machines at a fraction of the cost. The purpose of this research project is to explore this application arena: what kind of architectures are best (fastest)? How might they be automatically configured to self-organise? How might we build bridges between this new technology and a nascent user base? Industry has invested heavily - quite sensibly - in computing technology over the years, and if POETS is to become the disruptive technology we believe it to be capable of, we need to address a serious 'hearts and minds' issue for commercial uptake to ensue.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know