[Crib-list] TODAY: Speaker: OSKAR MENCER -- "Computational Research in Boston and Beyond Seminar" -- Friday, March 1, 2013 -- TIME: 12:00 Noon in Building 32, Room 141 (Stata Center) (fwd)

[Shirley Entzminger]

T O D A Y . . .

 		     COMPUTATIONAL RESEARCH in BOSTON and BEYOND SEMINAR


DATE:		FRIDAY, MARCH 1, 2013
TIME:		12:00 Noon
LOCATION:	Building 32, Room 141   (Stata Center)

Pizza and beverages will be provided at 11:45 AM outside Room 32-141


TITLE:		Multiscale Dataflow Computing

SPEAKER:	OSKAR MENCER   (CEO, Maxeler Technologies)


ABSTRACT:

Complexity of computation is a function of the underlying representation. We 
are extending this basic concept to consider representation of computational 
problems on the application level, the model level, the architecture level, 
arithmetic level and gate level of computation. In particular, the first step 
is to consider and optimize the discretization of a problem in time, space and 
value. Discretization of value is particularly painful, both in Physics where 
atomic discretization ruins many nice theories, and in computation, where most 
people just blindly use IEEE double precision floating point so they don't have 
to worry about details, until they do.  Multiscale Dataflow Computing provides 
a process by which one can optimize the discretization of time, space and value 
based on particular underlying computer architecture, and in fact, iterate the 
molding of the computer architecture and the discretization of the 
computational challenge.

The above methods have been able to achieve 10-50x faster computation per cubic 
foot and per Watt, resulting in less nodes per computation and therefore 
exponentially improved reliability and resiliency.  Results published by users 
worldwide include financial modeling (American Finance Technology Award for 
most cutting edge technology, 2011), commercial deployment in the Oil&Gas 
industry (see Society of Exploration Geophysicists meetings and reports), 
weather modeling (reducing time to compute a Local Area Model - LAM from 2 
hours to 2 minutes) and even sparse matrix solvers which cannot be 
parallelized, running 20-40x faster.

*******************************************************************************

Massachusetts Institute of Technology
Cambridge, MA


For more information about the Cribb Seminar, please visit...


 			http://math.mit.edu/crib