Abstract

The human visual system is extraordinary powerful, but it is not a perfect seeing device. In this talk I will use the example of binocular vision to explore visual processing. I will describe some of the biology of the binocular visual system, some of the limitations that the biology presents, and I will describe methods used to probe the relative importance of binocular vision, versus other sources of 3-D visual information. These methods allow us to predict where and when binocular vision provides a powerful source of 3-D information, and is therefore useful to inform the design and production of devices across a range of HCI applications.

Bio


Julie Harris is interested in visual perception, with particular interests in how binocular vision and eye movements are used for the perception of shape and depth and the control of action in 3-D space. Current projects include how binocular information is used for distance perception, how gaze patterns can be described in simple mathematical terms, and how we perceive motion in three dimensions.

Abstract

Despite Moore’s “law”, uniprocessor clock speeds have now stalled. Rather than using single processors running at ever higher clock speeds, it is common to find dual-, quad- or even hexa-core processors, even in consumer laptops and desktops. Future hardware will not be slightly parallel, however, as in today’s multicore systems, but will be massively parallel, with manycore and perhaps even megacore systems becoming mainstream. This means that programmers need to start thinking parallel. To achieve this they must move away from traditional programming models and development processes that offer parallelism as an bolted-on afterthought.

This talk introduces the idea of “paraforming”, a new approach to constructing parallel functional programs using formally-defined refactoring transformations.
We show how parallel programs can be built from a small number of primitive Haskell building blocks, and describe some new refactorings for Parallel Haskell that capture common parallel abstractions, such as divide-and-conquer and data parallelism using these building blocks. Using a paraforming approach, we are able to easily obtain significant and scalable speedups (up to 7.8 on an 8-core machine).

Continue reading