• Subscribe

Interview: Gerhard Wellein

"Young people need to understand that supercomputing is not just a single specific field," says Wellein. "They need to be interested in various aspects of the business, for example, the hardware, software layers, the application codes, numerical methods and finally the problem itself."

An accomplished author and a PhD holder in solid state physics, Gerhard Wellein will be delivering a keynote on Fooling the Masses with Performance Results: Old Classics & Some New Ideas, at the June International Supercomputing Conference (ISC'13) in Leipzig, Germany.

As a professor for computer science, Wellein teaches HPC techniques to students from computational science and engineering programs at the University of Erlangen-Nuremberg, Germany. He also leads the HPC group at the Erlangen Regional Computing Center, where he fulfils his passion for solving large, sparse eigenvalue problems. He recently talked to ISC's PR Manager, Nages Sieslack about his interesting life as a scientist and everything else he does outside the realm of supercomputing.

You hold a PhD in solid-state physics and you're in supercomputing. What inspired you to pursue a career in supercomputing?

As a post-graduate student I was looking for an interesting diploma thesis, and I decided to write about parallel numerical simulation on a 'supercomputer'. I had no clue about both areas, but they sounded interesting enough as topics.

Efficient programming of parallel systems fascinated me from the very beginning - actually at that time I was not allowed to use the Cray vector machines (which were reserved for the post docs), and had to cope with a Thinking Machines CM5. With every new generation of supercomputers (Intel Paragon, Cray T3D/E, SGI Origin/Altix, etc.) new problems had to be solved for parallel efficiency. I soon realized that physicists think in a way that helps to carry out efficient parallel numerical calculations. Together with my PhD advisor we addressed several interesting physics problems with the new classes of parallel computers, but at the end, all of my work about the parallel numerical simulation had to be put into the appendix of my dissertation - which was of the same size as the main part of my thesis. Fortunately things have changed since back then.

After my dissertation I went to Erlangen Regional Computer Center to look at this business from the other side of the table. There I had the opportunity to establish a group which actively works at the interface between application science, computer science and computer center, which continues to provide me a lot of fun.

Is there an individual that you admire most in high-performance computing?

There are so many excellent people in HPC addressing all the different aspects of it. If I have to name one individual, it is Seymour Cray, who is for sure one of the greatest on the architecture side. On the performance and programming side, I was influenced by Willi Schönauer from Karlsruhe, Germany, from very early on. Particularly, his book Scientific Supercomputing: Architecture and Use of Shared and Distributed Memory Parallel Computers provided a structured understanding of the effects of code optimization and parallelization, which is naturally very appealing to a physicist.

What do you do when you aren't busy thinking about parallel programming? Do you have any non-HPC-related hobbies?

First of all, I try to spend as much time as possible with my family, which consists of me, my wife and our four children. When there are some time slots available, I try to run several times a week. Running is in many aspects very similar to parallel programming. If you run the same distance every day, there are always different burdens on the way to the finish line. Sometimes the weather is great (compiler does a great job), but sometimes it starts to rain when there is still an hour left to the finish line (compiler cannot be convinced to produce reasonable code) and you hate every step that you take (you need to do assembly language programming yourself). But upon crossing the finish line, you forget about the pain.

While you teach HPC system to students and scientists, you also have four young children at home. How do you explain supercomputing to them?

This is a tough question, since my oldest kid is just nine years old. I travel multiple times a year to the US, which of course is understood by my children. Since we spent two vacations at Disneyworld, Orlando, they tell their friends that I work at Disneyland - at least when I am on travel. They are still too young to ask specific questions about my job. As my wife is a school teacher, they consider me as a teacher for older kids.

How did you develop your appreciation for science and logic?

I was already interested in math and physics when I was at elementary school and the interest stayed with me. There was no special incident that attracted me to those fields, but I never really focused on natural sciences, math and logic alone - I did also study economics for a while when I was a PhD student.

What do you think young people should do to enter the field of supercomputing?

Young people need to understand that supercomputing is not just a single specific field. They need to be interested in various aspects of the business, for example, the hardware, software layers, the application codes, numerical methods and finally the problem itself. This sounds complicated and exhaustive, but this is also a lot of fun and kind of ensures that working with supercomputers will never be boring. It also leads you to a career that allows you to meet many people, often from completely different backgrounds - which I personally find very exciting still today.

Join the conversation

Do you have story ideas or something to contribute?
Let us know!

Copyright © 2015 Science Node ™  |  Privacy Notice  |  Sitemap

Disclaimer: While Science Node ™ does its best to provide complete and up-to-date information, it does not warrant that the information is error-free and disclaims all liability with respect to results from the use of the information.


We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit ScienceNode.org — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on ScienceNode.org” containing a link back to the original article.
  4. The easiest way to get the article on your site is to embed the code below.