Dr N Filer
Applications accepted all year round
Competition Funded PhD Project (Students Worldwide)
About the Project
The current and almost certainly the next generation of multiprocessor hardware relies on wired connectivity between many core processors in a hierarchy of cores, chips, boards and racks. Currently the interconnect between all the components is achieved using high speed serial or bus connections. We believe there are many configurations where wireless communications will be either faster or lower in energy usage than the existing wired connections. Clearly there will also be lots of situations where wired connectivity should be preferred to wireless due to lower latency, higher concurrent channel utilization and noise resistance etc..
There a few existing studies of wireless networking on silicon systems. The aim of this research is to investigate these existing systems and to attempt to develop tools to aid hardware system architects in choosing suitable interconnect for proposed systems on silicon. For example, the Spinnaker system, the schools hardware brain simulator which currently uses chips with 20 ARM cores each to simulate massive numbers of neural circuits could act as an initial exemplar to be studied. For example, would a 64, 128, 256... core chip be more or less efficient if all or some inter-core connections and data was sent via wireless rather than through on-chip wiring?
We are very unlikely to be able to build silicon to test ideas in this field so the project will use analytical and simulation technologies to assess the characteristics of competing solutions.
Funding Notes
Candidates who have been offered a place for PhD study in the School of Computer Science may be considered for funding by the School. Further details on School funding can be found at: http://www.cs.manchester.ac.uk/study/postgraduate-research/programmes/phd/funding/school-studentships/.