Ruminations on High Performance Computing

Date: 
Friday, October 16, 2009 - 11:05am

*The Center for Control, Dynamical Systems, and Computation presents:*

*”Ruminations on High Performance Computing**”*

*Bruce Hendrickson*
Senior Manager for Computer Science & Mathematics – Sandia National Labs
*
Friday, October 9, 2009 4:00 – 5:00pm, HFH 1104**
*

*Abstract*:

After fifteen years of stasis, high performance computing is entering an era
of disruptive change. Several trends are driving this change. Traditional
applications are evolving rapidly, and new applications with different
computational needs are emerging. At the same time, computer architects are
facing new design constraints, yet the continued march of Moore’s Law
creates opportunities for radical innovation. Meanwhile, the field is in
urgent need of new abstractions and programming models. The combination of
these forces will raise enormous challenges in the coming years. But in
addressing these challenges, high performance computing will be enriched and
rejuvenated, and will contribute in unprecedented ways to science and
society.

*About the speaker:
*

My research interests include scientific computing, parallel algorithms
and combinatorics. Much of my work has been in Combinational Scientific
Computing – the development,
analysis and application of discrete algorithms for applications in
scientific computing. Some of my recent (and not so recent) projects include
the interplay between graph algorithms and computer architectures, support
theory – an algebraic approach to analyzing preconditioners, the Chaco graph
partitioning tool and the related Zoltan dynamic load balancer, and a range
of algorithmic activities in scientific and high performance computing.