Multiscale Dataflow Computing:
The Vertical Perspective

Oskar Mencer and Stephen P. Weston
Maxeler Technologies

In this talk we show a vertical perspective on high performance computing. While on the chip level, microprocessors look like the most efficient way to compute, once we evaluate efficiency on the architecture level, the more flexible approach we employ with multiscale dataflow computing allows us to further minimize communication and balance computation with communication components. Finally, on the system level, parallel programming for both multicore and dataflow based systems requires significant effort, and legacy software presents a major obstacle to change. We will show performance results from users at universities and industry.

To complete the picture, we present the application of Multiscale Dataflow Computing to complete scenario-based risk solutions.

Slides:

The slides for this presentation may be downloaded in two parts: Part 1 and Part2. The first part is fairly large (11MB); please let us know if you have difficultly downloading it.

About the speakers:

Oskar Mencer is founder, CEO and CTO of Maxeler Technologies. Oskar holds a first level degree of the Austrian Academy of Skiing (OSV) for skiing instructors, an undergraduate Computer Engineering degree from the Technion, and a PhD in Electrical Engineering (CSL) from Stanford University. Prior to founding Maxeler Technologies, Oskar Mencer was Member of Technical Staff at Bell Labs, leading the effort on "Computing with FPGAs" within the Computing Sciences Center. Besides his work with Maxeler, Oskar is also affiliated with the Computing Department at Imperial College London and holds a Consulting Professor position at the Geophysics department at Stanford University. Oskar's experience includes start-ups and more established companies such as DIGITAL, Rockwell and Hitachi (Tokyo).

Stephen Weston is currently Chief Development Officer at Maxeler. Prior to joining Maxeler, Stephen was managing director and global head of the Applied Analytics group at J.P. Morgan. The group is responsible for accelerating mathematical models of finance for trading and risk management. Stephen championed of the adoption, development and implementation of dataflow technology at J.P. Morgan. Prior to joining J.P. Morgan, Stephen also held senior quantitative and risk positions at several major investment banks. Stephen holds a PhD in mathematical finance from Cass Business School and is a director of risk management research at the Centre for Advanced Financial Engineering at Imperial College.

Contact information:

Oskar Mencer and Stephen P. Weston
Maxeler Technologies
London UK

http://www.maxeler.com