Computer Science Department Events
There will be a Colloquium on Thursday August 9 at 10:00 AM by David Keyes
Thursday, August 9, 2012 Constant Hall 1002TIME: 10:00 AMTITLE: 'High Performance Computing in Science and Engineering: the Tree and the Fruit' and 'The Missing Mathematics of Extreme Scale Simulation'SPEAKER: David Keyes, Columbia UniversityAbstract:
High Performance Computing in Science and Engineering: the Tree and the Fruit
Most people aware of the U.S. High Performance Computing and Communications Act of 1991 (Public Law 102-194, a.k.a. the "Information Superhighway Bill" or the "Gore Bill") note its transformative effects on the development of the information economy, the enhanced productivity of the overall economy, and the democratization of information, tracing a straight line to such diverse phenomena as Google, smart highways, and the Arab Spring. In fact, the high performance computing program, as implemented over the past 20 years by the 15 agencies of NITRD, has also transformed science and engineering, with high performance computing becoming, itself, a science. We reflect briefly not just on the fruit of HPC, but on the tree, itself, and on a host of "piggyback Moore's Laws".
Click here for Webcast and slides
The Missing Mathematics of Extreme Scale Simulation
An oft-quoted motivation for extreme computing is to "take the gloves off" with critical energy and environmental simulations: improve prediction by relaxing the decoupling, rolling out the full physics, cranking up the resolution, and quantifying the uncertainty. Meeting these objectives will indeed justify the daunting development, acquisition, and operation costs of the hardware. New hardware is, however, only one challenge and perhaps neither the highest risk nor the highest payoff. Much mathematics (and software) appears to be missing if the hardware is to be used at its potential. For instance, the promises of multiphysics simulation will not be realized in extreme-scale computational environments in the primary manner through which codes are coupled today, through divide-and-conquer operator splitting. Furthermore, today's successful, decoupled applications will need to be substantially rewritten after two decades of algorithm refinement in a period of programming model stability. While undertaking the latter, it may be natural to rethink the former.
Click here for slides.
|
