ὅδε οἶκος, ὦ ἑταῖρε, μνημεῖον ἐστιν ζῴων τῶν σοφῶν ἀνδρῶν, καὶ τῶν ἔργων αὐτῶν

DATAFLOW SUPERCOMPUTING FOR APPLICATION SPEEDUPS AND ENERGY SAVINGS



Project approved by Ministries of Sciences of Slovenia and Serbia
as a part of bilateral cooperation between two countries

Dates: January 2014 - December 2015

Project number: Project III 044006

Institutions:
from Slovenia: Univerza v Ljubljani, Fakulteta za elektrotehniko (Ljubljana)
from Serbia: Mathematical Institute SANU (Belgrade)

Project coordinators:

1. Sašo Tomažič, Fakulteta za elektrotehniko, Univerza v Ljubljani, Slovenia (email: saso.tomazic@fe.uni­lj.si)
2. Zoran Marković, Mathematical Institute SANU, Knez Mihailova 36, 11 000 Belgrade, Serbia (email: zoranm@mi.sanu.ac.rs)

Other project members:

from Slovenia

1. Niko Gamulin, Fakulteta za elektrotehniko, Univerza v Ljubljani, Slovenia
2. Anton Kos, Fakulteta za elektrotehniko, Univerza v Ljubljani, Slovenia
3. Sara Stančin, Fakulteta za elektrotehniko, Univerza v Ljubljani, Slovenia
4. Grega Jakus, Fakulteta za elektrotehniko, Univerza v Ljubljani, Slovenia
5. Jaka Sodnik, Fakulteta za elektrotehniko, Univerza v Ljubljani, Slovenia

from Serbia

1. Veljko Milutinović, School of Electrical Engineering, University of Belgrade, 11 000 Belgrade, Serbia
2. Jakob Salom, Mathematical Institute SANU, Knez Mihailova 36, 11 000 Belgrade, Serbia
3. Zoran Ognjanović, Mathematical Institute SANU, Knez Mihailova 36, 11 000 Belgrade, Serbia
4. Vladisav Jelisavčić, Mathematical Institute SANU, Knez Mihailova 36, 11 000 Belgrade, Serbia
5. Aleksandar Mihajlović, Mathematical Institute SANU, Knez Mihailova 36, 11 000 Belgrade, Serbia

Project goals: To establish the cooperation between researchers from Serbia and Slovenia in the field of applied dataflow supercomputing.
The objective of this project is to provide a set of algorithms, tools and needed skills for scalable and an order of magnitude faster data analytics
to scientific and industry worlds enabling them, with great reduction of energy consumption, to cope with Big Data (exascale data)
by utilizing native DataFlow Computing paradigm.
In the research and development process, special attention will be devoted to the following three methodological principles:

- algorithmic changes for better utilization of DataFlow architectures
- re-choreography of input data for just-in-time/just-in-place data streaming
- precision-related analysis