News and Announcements
Russian Academy of Sciences
Jack Dongarra and Vladimir Voevodin
Jack Dongarra recently received his certificate as a Foreign Member of the Russian Academy of Sciences (RAS). Established in 1724, the Academy is home to over 1,000 smaller institutions throughout Russia and houses a network of researchers in fields ranging from nuclear physics, applied mathematics, to philosophy and language.
When asked about the news of his induction into the RAS in 2016, Jack noted that “being elected to the academy is not only an honor but also another effective avenue for sharing what we learn from our experimental computer science work.” Congratulations, Jack!
Five UT Centers at SC19
Five computational science research centers from the University of Tennessee—the Bredesen Center, the Global Computing Laboratory, the Joint Institute for Computational Sciences, the Innovative Computing Laboratory, and Chattanooga’s SimCenter—will represent the University with their own booth (#575) at this year’s International Conference for High Performance Computing, Networking, Storage, and Analysis (SC19) on November 17–22 in Denver, Colorado. Stop by and say “Hello!”
ICL @ SC19 Dinner
The ICL @ SC19 dinner is set for Wednesday, November 20th at 6:30 p.m.
Henry’s Tavern (map)
500 16th Street, Unit 184B
Denver, CO 80202
720-399-8466
Respond to Joan Snoderly (jsnoderly@icl.utk.edu) by November 8th if you wish to attend.
Conference Reports
Brookhaven GPU Hackathon
Always on the move, Piotr Luszczek traveled to Upton, NY on September 23–27, where he served as a hacking mentor for the Brookhaven National Laboratory’s GPU Hackathon. The Upton event garnered around 40 attendees who were divided into teams working to port development codes to GPUs or further optimize their applications for the latest and greatest in GPU hardware—all with the help of a team mentor.
Piotr’s team worked on integrating NWChemEx with libraries from ICL and the Extreme-scale Scientific Software Development Kit (xSDK) with the aim of accelerating the application’s performance using NVIDIA GPUs.
This exercise is part of Piotr’s continued push to improve application integration with respect to xSDK and ICL libraries. The next GPU Hackathon is hosted by ORNL on October 21–25 at the Crowne Plaza hotel in Knoxville, TN. Of note: 2019 has seen a record 11 GPU Hackathons, and it seems like Piotr has participated in all of them.
The Editor would like to thank Piotr Luszczek for his contributions to this article.
SPEC19: HPC2020
The Standard Performance Evaluation Corporation’s 2019 workshop (SPEC19) was held on September 12–13, 2019 in Alexandria, VA and focused on benchmarking the next generation of leadership-class computers (hence “HPC2020”). Piotr Luszczek attended the workshop to outline ICL’s latest contributions to the field.
In presenting “Benchmarking: Doomed to Succeed or Simply Succeeding,” Piotr describes his experience in the design, implementation, and deployment/evaluation of 4 widely known benchmark efforts in which he was involved. Outlining the extensive adoption and longevity of these efforts, combined with the large volume of results and data, provided a relevant background on the community process involved in benchmarking and how this process (and experience) can be leveraged for the benefit of the SPEC HPC2020 effort.
According to Piotr, SPEC19 had plenty of informal opportunities to exchange ideas, and—even though there might have been differences of opinion as to which benchmark is the best—everybody was cordial. That’s always a plus. Familiar faces included David Bailey (HPCC), Bill Kramer (NCSA), Jack Wells (OLCF), and Rich Vuduc (Georgia Tech).
The Editor would like to thank Piotr Luszczek for his contribution to this article.
Recent Releases
heFFTe 0.1 Released
The Highly Efficient FFTs for Exascale (heFFTe) library was released on October 1st, 2019. heFFTe 0.1, the first release in ICL’s ECP-FFT effort, aspires to be a new and sustainable high-performance FFT library for exascale platforms that leverages the large investments in FFT software by the broader HPC community.
The heFFTe library has achieved significant acceleration of 3-D FFTs using GPUs, showing very good weak and strong scalability, owing to highly optimized GPU kernels and MPI communications.
heFFTe 0.1 is now available on FFT-ECP’s Bitbucket repo: https://bitbucket.org/icl/heffte/.
Interview
Sebastien Cayrols
Where are you from, originally?
I am a 33-year-old French guy born in Paris. I grew up in a small town close to Paris, and then I moved back to Paris for middle and high school.
Can you summarize your educational background?
I did not know what I wanted to do, but I was mainly interested in sciences. Following a teacher’s advice in high school, I started to really study Physics at the University of Paris-Sud XI.
During my Bachelor’s in Fundamental Physics, I took some programming courses, and I found it so interesting that I decided to quit physics and start again as a first-year student; it took me five more years to graduate as a “computer scientist” instead. And, during my last year, I had Laura Grigori as an instructor, and she then became my thesis advisor.
Where did you work before joining ICL?
I started my PhD working for INRIA, and I was primarily located at “Maison de la Simulation,” which is part of CEA Saclay. I was also a member of the ALPINE Team, which is a joint team between LJLL, Sorbonne University, and INRIA Paris. My main focus was redesigning algorithms to reduce the amount of communication.
Before defending my PhD, I received a postdoc position with Iain S. Duff at RAL, STFC in the United Kingdom. I then became part of a European project called NLAFET, and I was working on sparse-direct solvers.
How did you first hear about the lab, and what made you want to work here?
I first heard about ICL during my last year as a (an under?)graduate. Marc Baboulin, who was one of my professors, had invited a famous American guy to give a talk at LRI, Orsay, and Marc informed us that we MUST attend. This “American guy” turned out to be Jack. To be honest, I was new to HPC at the time, and I discovered so many things through his talk.
So, a few years later, when I joined NLAFET, I met Jack, again, and he invited me to Knoxville. Finally, I discussed the invitation with Marc, Mawussi, and Florent, and they told me how great it is to be at ICL, and I said why not go to Tennessee if my wife agrees?
What is your focus here at ICL? What are you working on?
I would say that I am a sparse guy, in the mathematical sense, because of what I have done during my PhD, and I have the feeling that my skills are like a gruyere, which is Swiss cheese with so many holes. Being here should allow me to partly remedy this.
I want to do GPU programming, machine learning, and efficient programming just to cite a few. To this end, I am involved in some very interesting projects: PEEKS, FFT, Convex Optimization, and AsyncIS.
What are your interests/hobbies outside of work?
I have a 3-year-old son, and my wife and I are waiting for a second child due in October! 🙂
Otherwise, nothing as exciting as kids I am afraid, though I do enjoy traveling, series, mangas, games, and cooking.
Tell us something about yourself that might surprise people.
I never did want to learn English. I spent years at school avoiding it. I was even targeting a job where English is not necessary. So now the question is: what happened? I still do not know. I guess doing a PhD pushed me to change my mind (actually, I had no other choice).
If you weren’t working at ICL, where would you like to be working and why?
I have no idea.
