Univerza na Primorskem Fakulteta za matematiko, naravoslovje in informacijske tehnologije
SI | EN

petek, 20. marec 2015 Seminar MARA

V ponedeljek, 23. marca 2015, bo ob 16.00 uri v prostorih Fakultete za matematiko, naravoslovje in informacijske tehnologije Univerze na Primorskem, Glagoljaška 8, Koper predavanje v okviru skupnega SEMINARJA ZA MATEMATIČNE IN RAČUNALNIŠKE ZNANOSTI Oddelka za matematiko in Oddelka za Informacijske znanosti in tehnologije UP FAMNIT, Oddelka za matematiko in Oddelka za Informacijske znanosti in tehnologije UP IAM, Oddelka za matematiko in računalništvo UP PEF ter Oddelkov za matematiko in teoretično računalništvo IMFM.

RAČUNALNIŠKI SEMINAR

Prostor: FAMNIT-1-MP2 ob 16:00

LECTURER: Prof. Veljko Milutinovic,

Fellow of the IEEE
Life Member of the ACM
Member of Academia Europaea
Member of the Serbian Academy of Engineering
Member of the Scientific Advisory Board of Maxeler
Member of the Scientific Advisory Board of Moskowitz-Jacobs

TITLE: DataFlow SuperComputing for BigData Analytics

ABSTRACT:
This presentation analyses the essence of DataFlow SuperComputing,  defines its advantages and sheds light on the related programming  model. DataFlow computers, compared to ControlFlow computers, offer  speedups of 20 to 200 (even 2000 for some applications), power  reductions of about 20, and size reductions of also about 20. However,  the programming paradigm is different, and has to be mastered. The  talk explains the paradigm, using Maxeler as an example, and sheds  light on the ongoing research in the field. Examples include  DataEngineering, DataMining, FinancialAnalytics, etc. A recent study  from Tsinghua University in China reveals that, for Shallow Water  Weather Forecast, which is a BigData problem, on the 1U level, the  Maxeler DataFlow machine is 14 times faster than the Tianhe machine,  which is rated #1 on the Top 500 list (based on Linpack, which is a  smalldata benchmark). Given enough time, the talk also gives a  tutorial about the programiing in space, which is the programming  paradigm used for the Maxeler dataflow machines (establishe din 2014  by Stanford, Imperial, Tsinghua, and the University of Tokyo).

Vabljeni!