[Colloquium] Staff] Computer Science Colloquium Talk

[Donna Brooms]

The University of Chicago Computer Science Department PRESENTS
 
COLLOQUIUM SPEAKER
 
Friday, February 1, 2013
2:30 p.m.
Ryerson 251
 
David E. Keys
Professor, Applied Mathematics and Computational Science
Director, Strategic Initiative in Extreme Computing
King Abdullah University of Science and Technology (KAUST)
 
TITLE:
Large-­‐scale Simulation in   Science   and Engineering  Digesting the Fruit, Replanting the Fields
 
ABSTRACT:
 
High performance computing has transformed science and engineering, and become, itself, a science The Scientific Discovery through Advanced Computing (SciDAC) program of the U.S. Department of Energy, among other campaigns of the past decade, has raised expectations and standards for predictive simulation as a tool of discovery, design, and decision support. The merger of the “third paradigm” of simulation for modeled systems with the “fourth paradigm” of  data analytics independent of physical models has opened up   exciting   prospects o  data assimilation and uncertainty quantification. These developments ride the dual trends of three orders of magnitude per decade in increased aggregate performance and three orders of magnitude per decade in decreased cost of installed flop/s. However, diverging exponentials I hardware subsystem performance no threaten to  interrupt the  steady progress in Computational Science and Engineering and require rethinking of models and reimplementation of algorithms. Much mathematics and software appears to be missing if emerging hardware I to be use near its potential, since our existing scientific computing code  base has  been assembled with a premium on squeezing out as many floating point operations as possible and improving the execution rate of those that remain. Instead, for reasons of energy efficience   and system acquisition cost, algorithms must now focus on squeezing out synchronizations, memory footprint, and memory transfers, including copies. High concurrency and power-­‐efficient design of  the individual core  put opposite pressures on algorithms: requiring simultaneously greater data locality and greater freedom to redistribute data and computation. After decades of  algorithm refinement during   period of programming model stability, new programming models and new capabilities must be developed simultaneously. We extrapolate current trends and describe new direction
 
 
Host: Prof. L.Ridgway Scott
 
*Refreshments will be served following the talk at 3:30 in Ryerson 255

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20130122/3431c57b/attachment.htm