Gibbons Lecture Series: Attached processors for real-time applications
Presented in association with the NZCS
Speaker: Associate Professor John Morris, Dept of Computer Science & Dept of Electrical and Computer Engineering, The University of Auckland
When: Refreshments at 5.30pm, lecture starts at 6.00pm.
John Morris has a PhD in optical spectroscopy from the University of Sydney, but after an encounter with a 30kV pulsed laser power supply, moved into the much safer computing field - encouraged by the fact that modern processors run on a mere (and touchable) 1.8V now. He has taught in Melbourne, Tasmania, Western Australia, ChungAng (Seoul), KyungPook (Daegu) and Auckland and had post-docs in NRC Ottawa and Tokyo. His main research interest is computer architecture and reconfigurable systems and is returning to his starting point in optics with a focus on intelligent vision systems. The high-resolution real-time stereo vision systems in the laboratory at Auckland are a first step in a grand goal to emulate human eye-brain capabilities.
Synopsis: Although modern single processor computers can easily perform a billion calculations per second, they still cannot even match the capabilities of our vision systems! However, specialized processors may be attached to a conventional computer to enhance its power. Such processors include reconfigurable logic chips - in which the computing circuits can be changed at will - and the very powerful graphics processing units (GPUs) which are of particular interest nowadays since there's one in almost every desktop computer. We'll look at harnessing a GPU's power to solve some problems that require enormous computing power - one example is human genome matching. Another is a system, developed in our Photogrammetry Laboratory, which can see a 3D world using stereo vision in the same way that our brain uses two eyes to work out the distance to objects that it can see. Such systems will allow computers to take over many tasks that, for humans, are dangerous, boring or error-prone. But GPUs were designed for a specific application - rendering scenes rapidly for electronic games - and thus are more appropriate for some applications than for others.