News and Announcements

SC16 Student Cluster Competition

Registration for the SC16 Student Cluster Competition closes on April 22, 2016. If any ICL undergrad students are interested in putting your HPC knowledge and skills to the test, now is your chance to compete. Unlike previous years, the 2016 competition involves “the SCC Reproducibility Initiative,” in which students will be challenged to reproduce a paper rather than run prescribed datasets. For the first time in the competition, students will be responsible for replicating results from an award-winning paper that was presented at a previous SC conference (i.e., SC15).

The SCC is a real-time, non-stop, 48-hour competition where students assemble small cluster computers on the SC16 exhibit floor and race to complete a real-world workload across a series of applications and impress HPC industry judges. Prior to the competition, teams work with their advisor and vendor partners to design and build a cutting-edge cluster from commercially available components that does not exceed a 3120-watt power limit (26-amp at 120-volt), and work with application experts to tune and run the competition codes.

Click here for more information.

Conference Reports

PPoPP’16

ICL’s Thomas Herault ventured to Barcelona, Spain for the 21st ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming (PPoPP’16), held on March 12th-16th. PPoPP is a forum for bleeding edge work on all aspects of parallel programming, and is particularly interested in work that addresses new parallel workloads, techniques, and tools that attempt to improve the productivity of parallel programming.

For Thomas’s part, he co-hosted a tutorial with ICL alum and frequent visitor Yves Robert. The tutorial, “An overview of fault-tolerant techniques for HPC,” presented a comprehensive survey of the techniques proposed to deal with failures in high performance computing systems. The tutorial’s overarching goal was to provide the participants with a clear picture by expounding on the techniques and how they work, and provide an overview of the basic tools and knowledge necessary for quantitative and qualitative evaluation.

PAPI All Hands Meeting

On March 23-24th, ICL hosted the first PI meeting for the NSF funded PAPI-EX project. Anthony Danalis, Heike Jagode, Phil Mucci (Minimal Metrics), Vince Weaver (U of Maine), and Asim YarKhan all participated in the meeting where they discussed progress in PAPI-EX and goals for next year.

PAPI-EX reached several milestones in Year 1, including creating a new “powercap” component that can provide read and write access to the information and controls exposed via the Linux powercap interface. The new component supports measuring and capping power usage on recent Intel architectures, and is the first case where PAPI is writing information to a counter as well as reading the data. Users are encouraged to access and test the new component, which is currently available in the PAPI Git repo, and will be officially released in the next PAPI tar ball.

The papiex tool, developed by Phil Mucci at Minimal Metrics, was also released and will be integrated into an upcoming stable release of PAPI.

Upcoming goals for the PAPI team include developing—in collaboration with Intel—PAPI support for KNL. The PAPI team is also experimenting with micro-kernels to validate native events as part of a new Counter Inspection Toolkit (CIT).

Interview

Terry Moore Then

Terry Moore

Where are you from, originally?

I was born and raised in Ft. Wayne, Indiana. Sue Dongarra is also from Ft. Wayne. In fact, we both grew up on the north side of town, maybe 10 miles apart, but of course not at the same time. It’s a small world after all.

Can you summarize your educational background?

I got my undergraduate degree at Purdue. I took the engineering curriculum my first two years because that was sort of the family trade—both my father and brother were chemical engineers. But unfortunately, I swerved off the designated track and went into Philosophy. My extended family found such an impractical choice incomprehensible, but made themselves feel better by occasionally chanting “Law school! Law school!” No such luck. I solidified my position as the family ne’er do well by going to graduate school in Philosophy at UNC, Chapel Hill, where I got both a Masters and (eventually) a Ph.D.

Where did you work before joining ICL?

I started at UTK in 1992 as the Director of Computing and Communications at the College of Nursing. The dean of the College, Joan Uhl Pierce, brought me with her from UNC, where she had been Associate Dean at the School of Nursing. I started my IT career at UNC around 1988. I was basically self-taught in the field, but hey, it was the late 80’s and early 90’s, the Internet was just beginning to roar, and there were lots of people who were winging it, like me. Fortunately, the Dean liked the idea of having someone with a Ph.D. doing basic sys admin stuff and desktop support. In late 1995, I went to work for Network Services as the faculty liaison. Network Services had excellent technical people, but no one who really understood academia very well or who could talk to faculty about their IT and networking needs. I came to ICL from Network Services.

How did you first hear about the lab and what made you want to work here?

I first met Jack as a result of a grant I wrote to Sun Microsystems for the campus CIO, which brought a “Sun Site” focused on Java to UTK. Java was a new thing at the time, and Jack sent an email to my boss asking about it. Jack listened to me very politely, but I was wearing a tie and I suspect I came across as an administrative dork of some variety. Fortunately, he seemed to have forgotten all about that when, about a year and a half later, he hired me as part of a package deal that also brought Micah Beck into ICL. We had written some successful NSF proposals together for Network Services, but the two of us had gotten “sideways” with the then CIO—a different one, whose head rolled shortly thereafter—and we had to flee the administrative side of the house. Micah convinced me that ICL would be an excellent place to work, and it seemed like a good time to roll the dice.

What is your primary role here at ICL? What are you working on?

ICL requires quite a bit of funding to keep going, and that means writing good proposals to government agencies that succeed at a relatively high rate. I usually describe the proposal development process as “team competitive writing.” I help to organize and manage the execution of such efforts. Developing proposals can be pretty strenuous, but because it is such a collaborative thing, and generally requires the cultivation and growth of new ideas, it can be very satisfying when things go well. Also, there’s the money.

What are your interests/hobbies outside of work?

Though I didn’t become a professional philosopher, I’ve maintained my interest in Philosophy, which you might say has become my hobby. That means, basically, that I spend much of my spare time doing a lot of reading and a little writing about things that are related, in one way or another, to my intellectual interests. I’m also an avid walker.

Tell us something about yourself that might surprise people.

When I was young I wanted to become a professional magician. I focused mostly on close up card magic, and worked pretty hard at it for several years. I impressed my bride to be with a card trick at a party, and the rest is history. Amazingly, she seems to have been far less impressed with the tricks I’ve done since we were married.

If you weren’t working at ICL, where would you like to be working and why?

To be honest, this question is more appropriate for a person in a much earlier phase in their career. At the stage I’ve reached, I’m not much inclined to think about working somewhere else. As I hinted above, I feel very fortunate to have found a place at ICL, and very privileged to have had such a tremendous group of colleagues and collaborators to work with over the years. I plan to continue to enjoy the good fortune I’ve had; contemplating alternatives seems counter productive to that plan.

Recent Papers

  1. Valero-Lara, P., J. Dongarra, A. Haidar, S. D. Relton, S. Tomov, and M. Zounon, A Standard for Batched BLAS Routines , Paris, France, 17th SIAM Conference on Parallel Processing for Scientific Computing (SIAM PP16), April 2016.  (1.93 MB)
  2. Abdelfattah, A., A. Haidar, S. Tomov, and J. Dongarra, Cholesky Factorization on Batches of Matrices with Fixed and Variable Sizes , San Jose, CA, GPU Technology Conference (GTC16), Poster, April 2016.  (480.51 KB)

Recent Conferences

  1. MAR
    Thomas Herault
    Thomas
    Thomas Herault
  2. MAR
    Copper Mountain Conference Denver, Colorado
    Ichitaro Yamazaki
    Ichitaro
    Ichitaro Yamazaki
  3. MAR
    PAPI-EX PI Meeting Hilton, Tennessee
    Anthony Danalis
    Anthony
    Asim YarKhan
    Asim
    Heike Jagode
    Heike
    Phil Mucci
    Phil
    Anthony Danalis, Asim YarKhan, Heike Jagode, Phil Mucci
  4. APR
    Ahmad Abdelfattah Ahmad
    Ahmad
    Azzam Haidar
    Azzam
    Stanimire Tomov
    Stan
    Ahmad Abdelfattah Ahmad, Azzam Haidar, Stanimire Tomov
  5. APR
    -
    SIAM PP Paris, France
    Aurelien Bouteiller
    Aurelien
    George Bosilca
    George
    Ichitaro Yamazaki
    Ichitaro
    Aurelien Bouteiller, George Bosilca, Ichitaro Yamazaki

Upcoming Conferences

  1. MAY
    Piotr Luszczek
    Piotr
    Piotr Luszczek
  2. MAY
    TESSE Meeting New York, New York
    Amina Guermouche
    Amina
    George Bosilca
    George
    Thomas Herault
    Thomas
    Amina Guermouche, George Bosilca, Thomas Herault
  3. MAY
    Amina Guermouche
    Amina
    Aurelien Bouteiller
    Aurelien
    George Bosilca
    George
    Ichitaro Yamazaki
    Ichitaro
    Thomas Herault
    Thomas
    Amina Guermouche, Aurelien Bouteiller, George Bosilca, Ichitaro Yamazaki, Thomas Herault
  4. MAY
    -
    Ahmad Ahmad
    Ahmad
    Azzam Haidar
    Azzam
    Hartwig Anzt
    Hartwig
    Jack Dongarra
    Jack
    Jakub Kurzak
    Jakub
    Mark Gates
    Mark
    Piotr Luszczek
    Piotr
    Stanimire Tomov
    Stan
    Terry Moore
    Terry
    Ahmad Ahmad, Azzam Haidar, Hartwig Anzt, Jack Dongarra, Jakub Kurzak, Mark Gates, Piotr Luszczek, Stanimire Tomov, Terry Moore
  5. MAY
    George Bosilca
    George
    Piotr Luszczek
    Piotr
    George Bosilca, Piotr Luszczek
  6. MAY
    Ahmad Abdelfattah Ahmad
    Ahmad
    Ahmad Abdelfattah Ahmad

Recent Lunch Talks

  1. MAR
    4
    Hartwig Anzt
    Hartwig Anzt
    Solving Sparse Linear Systems on GPUs - The Good, the Bad, and the Ugly
  2. MAR
    11
    Ichitaro Yamazaki
    Ichitaro Yamazaki
    Preconditioning a Communication-avoiding Krylov solver PDF
  3. MAR
    18
    Tim Davis
    Tim Davis
    Texas A&M University
    Sparse Matrix Algorithms: Combinatorics + Numerical Methods + Applications PDF
  4. MAR
    18
    Sanjay Ranka
    Sanjay Ranka
    University of Florida
    A Genetic Algorithm Based Approach for Multi-objective Hardware/Software Co-optimization PDF
  5. MAR
    24
    Phil Mucci
    Phil Mucci
    Minimal Metrics
    Systems Performance @ Sandia PDF
  6. APR
    1
    Ahmad Ahmad
    Ahmad Ahmad
    On the Development of Variable-Size Batched Computation for Heterogeneous Parallel Architectures PDF
  7. APR
    8
    Piotr Luszczek
    Piotr Luszczek
    Search Space Description, Generation, and Pruning System for Autotuners PDF
  8. APR
    15
    Miro Stoyanov
    Miro Stoyanov
    ORNL
    Resilient Solvers for Partial Differential Equations PDF
  9. APR
    22
    Chongxiao Cao
    Chongxiao Cao
    Fault Tolerant Design for a Task-based Runtime PDF
  10. APR
    29
    Wei Wu
    Wei Wu
    Accelerator Integration with Programming Models PDF

Upcoming Lunch Talks

  1. MAY
    4
    Yaohung Tsai
    Yaohung Tsai
    AlphaGo: The Go AI from Google DeepMind
  2. MAY
    6
    Oleg Shylo
    Oleg Shylo
    Department of Industrial & Systems Engineering at UTK
    Scalable Communication for Parallel Optimization
  3. MAY
    13
    George Bosilca
    George Bosilca
    PaRSEC - Yet another runtime? PDF
  4. MAY
    20
    Iain Duff
    Iain Duff
    the Numerical Analysis Group at the Scientific Computing Department of the Science and Technology Facilities Council (UK)
    Scalability of Sparse Direct Codes PDF
  5. MAY
    27
    Azzam Haidar
    Azzam Haidar
    Heterogeneous Computation: The Current Challenge

People

  1. Blake Haugen
    Blake Haugen successfully defended his PhD on March 25th. Congratulations, Blake!