Technological disruption has periodically re-invigorated the field of
high performance computation, opening up new application domains and
problem solving algorithms, together with novel software techniques
for implementing them.
A recent disruption, the Graphical Processing Unit (GPU), has brought
high performance parallel processing to everyone's finger tips. However,this opportunity comes at the cost increased application development
costs associated with managing hardware specific implementation details.
In this talk I will present results from various GPU-enabled
applications developed in my research group, discussing the
opportunities offered by the technology along with its limitations.
I'll conclude by outlining new research directions currently under
investigation.
|