Besides their original purpose as graphical processing units, GPUs are increasingly deployed in applications that require high computational performance. Moreover, in the context of high performance computing (HPC) one is observing the trend to deploy general purpose GPUs in order to enable the transit from petascale computing to exascale computing. Nonetheless, programming and constructing such architectures remains significant challenge and often a significant bottlenecks for many companies due to its highly specialized nature.
Tom Ashby studied for both
his undergraduate degree and PhD at the University of Edinburgh. The PhD was on the topic of
using high level languages for scientific
computation, and the work was carried out in cooperation with the Particle Physics group. After completing his
studies he did a post-doc on simulating
lightweight hardware for cache coherence
on shared memory platforms, and then took a position at IMEC.
At IMEC he has worked on program transformation tools to support practical combined functional and loop parallelism for software on multicore embedded systems. After the results of that project were transferred to a large Japanese conglomerate, he moved to working on hyperspectral camera system design and image processing using machine learning techniques. He is currently part of the Flanders Exascience lab (http://www.exascience.com/) working with Intel on the performance and reliability challenges for scientific simulation on exascale machines, and the computational demands of future life sciences and medical workloads.