An Introduction to Data-flow Computing with Examples

Datum

18.03.2015

Zeit

15:00 - 16:00

Sprecher

Dr. Goran Dimic

Zugehörigkeit

University of Belgrade Institute Mihailo Pupin

Sprache

en

Hauptthema

Elektro- u. Informationstechnik

Andere Themen

Informatik, Elektro- u. Informationstechnik

Host

Prof. E. Jorswieck

Beschreibung

This presentation provides an introduction to data-flow computing, shows its advantages and limitations, and describes the related programming model. For certain classes of algorithms and large data sets, data-flow computers, compared to control-flow computers, enable trade-off of the following benefits: speed-ups of 20 to 200 times, power reductions of about 20 times, and size reductions of about 20 times. However, the data-flow programming paradigm is different, and has a steep learning curve. The talk explains the paradigm, using Maxeler as an example, and sheds light on the ongoing research in the field. Examples of data-flow applications include seismic imaging, genetic mapping, weather simulation, etc. The talk also briefs on the programming in space, which is the programming paradigm used for the Maxeler data-flow machines. Some limitations of the data-flow paradigm are shown through a brief discussion of data-flow implementation of matrix multiplication.

Letztmalig verändert: 12.03.2015, 17:04:49