DataFlow Supercomputing for Big Data Deep Analytics

Wednesday February 27, 2019
Location: CIC 4th floor Panther Hollow Conference Room
Time: 4:30PM-6:00PM

This talk and the accompanying hands-on tutorial analyze the essence of dataflow supercomputing, define its advantages, and shed light on the related programming model. The dataflow paradigm, compared to the control flow paradigm, offers the potential for: (a) speedups of at least 10x to 100x and sometimes much more (depending on application characteristics), (b) better precision (depending on the characteristics of the optimizing compiler, operating system, etc.), (c) a power reduction of at least 10x, and (d) a hardware size reduction of well over 10x. Our Ultimate DataFlow supercomputing approach achieves all of these benefits, but its programming paradigm is different, and has to be mastered to fully take advantage of the benefits.

This presentation explains Ultimate DataFlow and its programming paradigm, using the Maxeler programming model as an example, and sheds light on ongoing research in the field by my group and others. The presentation concludes with a summary of notable applications of the Maxeler work, such as emulating quark- related processes, tensor calculus for emulating QuasiCrystals, financial applications used by large banks and trading exchanges, mathematical algorithms, image processing, and ML/AI.

Prof. Veljko Milutinovic received his PhD from the University of Belgrade in Serbia, spent about a decade on various faculty positions in the USA (mostly at Purdue University and more recently at the Indiana University in Bloomington), and was a co-designer of DARPA's first GaAs RISC microprocessor and first GaAs Systolic Array. Later, for about two decades, he taught and conducted research at the University of Belgrade, in EE, MATH, BA, and PHYS/CHEM. Now he serves as the Senior Advisor to Maxeler Technologies in London, and the Chairman of the Board of IPSI Belgrade. His research is mostly in data mining algorithms and dataflow computing, with an emphasis on mapping of data analytics algorithms onto fast energy efficient architectures. For 10 of his books, forewords were written by 10 different Nobel Laureates with whom he cooperated on his past industry sponsored projects.