News and Announcements

NICS HPC Seminar Series

The National Institute for Computational Sciences invites you to a Seminar Series on High Performance Computing, every Tuesday and Thursday from 2:10pm to 3:10pm in Claxton 233, starting on Tuesday October 1st. This is a joint effort between different leadership organizations (NICS, JICS, OLCF, ICL) to increase HPC awareness within the academic community.

Different topics will be introduced starting with the most basic and building up to more advanced topics in HPC. No registration is required for the seminar.

Calendar of topics to be covered in October:

Date Title
1 Supercomputing in the present. Why do we need it?
3 Overview on NICS/JICS/UTK/Keeneland/XSEDE and DOE/OLCF HPC resources
8 How/Where/When to apply for an allocation/account (DD/XSEDE/INCITE)
10 Intro to HPC and Supercomputing
15 How to login and use your account (OTP/GSISSH/XSEDE Portal/Science Gateway)
17 Fall Break
18 Intro to the Cray XC30 (DARTER)
22 HPC in Physics
24 Intro to Unix/Linux
29 HPC in Chemistry
31 Intermediate unix

ICL on LinkedIn

ICL is now on LinkedIn! Follow this link to find us on LinkedIn and connect with your fellow ICLers on the professional social network.

Conference Reports

PPAM 2013

PPAM1

Kicking off a busy month of conferences and workshops, ICL’s Jack Dongarra and Piotr Luszczek were both keynote speakers at the 10th International Conference on Parallel Processing and Applied Mathematics (PPAM’13), held in Warsaw, Poland on September 8 – 11. PPAM 2013, the tenth conference in the series, covered topics in parallel and distributed processing, including theory and applications, as well as applied mathematics.

Jack’s talk gave an overview of HPC trends and challenges, including a discussion of the influence of the HPL benchmark and how alternatives, which expose many aspects of a machine’s performance, not just FLOP/s, are going to be useful metrics as HPC marches on to Exascale.

Piotr’s talk outlined the power constraints faced by current HPC systems and how that will be compounded by the pursuit of Exascale. Another challenge described by Piotr is the time required to run an HPL benchmark, or Time-to-Run, which would take 5.5 days on an Exascale machine, using an incredible amount of power along the way.

CSCS-USI Autumn School

On September 14th and 15th, ICL’s Stan Tomov traveled to Lugano, Switzerland to give a tutorial on MAGMA at the Swiss National Supercomputing Center’s USI Autumn School, a 2-day intensive workshop on the use of numerical libraries for Graphical Processing Units (GPUs).

GPUs offer huge potential performance, but their programming can be challenging. One way to attain the promised performance is to invoke routines from libraries, like MAGMA, which have been specially designed and optimized for GPUs.

Stan’s tutorial was given in two parts, and provided a general overview of the MAGMA library, its advantages and uses, followed by some specific instructions on installation and functionality. PDFs of Stan’s tutorial are linked below.

magma_tutorial_1

MAGMA: Part 1

magma_tutorial_2

MAGMA: Part 2

 

Algorithms and Scheduling Techniques for Exascale Systems

Group

The Algorithms and Scheduling Techniques for Exascale Systems seminar, hosted by Schloss Dagstuhl – Leibniz Center for Informatics, brought in HPC specialists from all over the world to discuss problems and possible solutions for challenges in application design, fault-tolerance, and power management of Exascale machines.

Several ICLers attended the seminar in Wadern, Germany. George Bosilca gave a talk about using runtimes (like PaRSEC) to ease the run up to Exascale. Aurelien Bouteiller discussed enabling the production deployment of advanced fault tolerance techniques in MPI. In Anthony Danalis’s talk, he asked the question: “Does cache matter?” Yes, as it turns out, it does. Thomas Herault, also at the seminar, gave a talk on assessing the impact of ABFT & Checkpoint composite strategies.

There were also several ICL alumni and frequent collaborators at Dagstuhl, including Yves Robert, Henri Casanova, Julien Herrmann, and Emmanuel Jeannot.

EuroMPI

EuroMPI2013_group_photo

This year’s EuroMPI conference was held in Madrid, Spain on September 15 – 18. Jack Dongarra served as general chair of the conference, and the meeting focused on the efficient implementation of MPI, typically on high-performance computing platforms, benchmarking and tools for MPI, short-comings and extensions of MPI, parallel I/O and fault tolerance, as well as parallel applications using MPI.

Interview

Thananon “Arm” Patinyasakdikul Then

Thananon “Arm” Patinyasakdikul

Where are you from, originally?

I’m from Nakhonratchasima, Thailand. The name is very difficult to pronounce, but not as difficult as my name! The city also has a shorter nick name, “Korat.” If you don’t know already, Thailand is a place of natural beauty, full of charming seas and great mountains!

Can you summarize your educational background?

I attended Rajsima Wittayalai High School in my hometown, and then pursued my bachelor’s degree in electrical engineering from Suranaree University of Technology. Upon graduation, I decided to pursue my PhD at UT.

Tell us how you first learned about ICL.

At Suranaree, one of my professors was an ICL alumni. His name is Thara and I even saw him in a several ICL retreat photos posted in the break room. He mentioned ICL in his class when we were learning about the MPI standard.

What made you want to work for ICL?

When I worked closely with Thara, he shared a lot of his experiences about ICL. He mentioned all of the skill and background he gained from ICL, and how professional and kind ICLers were. I would like to be like him, follow in that path, and learn as much as I can from ICL.

What are you working on while at ICL?

I’m working on the OpenMPI project on George’s team. I’m currently trying (and failing) to overlap the communication between MPI processes.

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

I was always passionate about Google. I read a lot about it and talked to some people who work there. That place is No.1 in my mind. It would be nice to land a job there after I graduate.

What are your interests/hobbies outside work?

I do professional photography. I really like taking photos, but unfortunately I couldn’t fit my camera and lens in my bag on the trip here. You can see some of my work here: http://blog.9arm.in.th.

I also do game casting, where I commentate the game for the audience. It’s like a football commentator. I was a commentator in a major game tournament in Thailand. I also run a small YouTube channel where I keep videos of me commentating some games (in my language). I have around 750 subscribers and 200k+ views across my videos!

Tell us something about yourself that might surprise people.

When I was in high school, I formed a team to develop computer games. One of the games won national and international awards! The game genre is RPG. My team was in every national newspaper and also received some national television coverage!

Recent Papers

  1. Turchenko, V., G. Bosilca, A. Bouteiller, and J. Dongarra, Efficient Parallelization of Batch Pattern Training Algorithm on Many-core and Cluster Architectures,” 7th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems, Berlin, Germany, September 2013.  (102.51 KB)
  2. Dongarra, J., M. Gates, A. Haidar, Y. Jia, K. Kabir, P. Luszczek, and S. Tomov, Portable HPC Programming on Intel Many-Integrated-Core Hardware with MAGMA Port to Xeon Phi,” PPAM 2013, Warsaw, Poland, September 2013.  (284.97 KB)
  3. Mattson, T., D. Bader, J. Berry, A. Buluc, J. Dongarra, C. Faloutsos, J. Feo, J. Gilbert, J. Gonzalez, B. Hendrickson, et al., Standards for Graph Algorithm Primitives,” 17th IEEE High Performance Extreme Computing Conference (HPEC '13), Waltham, MA, IEEE, September 2013. DOI: 10.1109/HPEC.2013.6670338  (108.86 KB)
  4. Haidar, A., P. Luszczek, J. Kurzak, and J. Dongarra, An Improved Parallel Singular Value Algorithm and Its Implementation for Multicore Hardware,” University of Tennessee Computer Science Technical Report (also LAWN 283), no. ut-eecs-13-720: University of Tennessee, October 2013.  (1.23 MB)
  5. Faverge, M., J. Herrmann, J. Langou, B. Lowery, Y. Robert, and J. Dongarra, Designing LU-QR hybrid solvers for performance and stability,” University of Tennessee Computer Science Technical Report (also LAWN 282), no. ut-eecs-13-719: University of Tennessee, October 2013.  (4.11 MB)
  6. Aupy, G., A. Benoit, T. Herault, Y. Robert, and J. Dongarra, Optimal Checkpointing Period: Time vs. Energy,” University of Tennessee Computer Science Technical Report (also LAWN 281), no. ut-eecs-13-718: University of Tennessee, October 2013.  (440.13 KB)
  7. Yamazaki, I., T. Dong, R. Solcà, S. Tomov, J. Dongarra, and T. C. Schulthess, Tridiagonalization of a dense symmetric matrix on multiple GPUs and its application to symmetric eigenvalue problems,” Concurrency and Computation: Practice and Experience, October 2013.  (1.71 MB)

Recent Lunch Talks

  1. SEP
    6
    Yves Robert
    Yves Robert
    On the Combination of Silent Error Detection and Checkpointing PDF
  2. SEP
    13
    Jakub Kurzak
    Jakub Kurzak
    Parallel Ultra Light Systolic Array Runtime PDF
  3. SEP
    20
    Piotr Luszczek
    Piotr Luszczek
    Energy and Power Consumption Trends PDF
  4. SEP
    27
    Dan Terpstra
    Dan Terpstra
    Small Scale Water Treatment in Developing Countries PDF
  5. OCT
    4
    Aurelien Bouteiller
    Aurelien Bouteiller
    Multi-criteria Checkpointing Strategies: Response-time vs Resource Utilization PDF
  6. OCT
    11
    Alisa Meador
    Alisa Meador
    UTK Center for International Education
    International SOS
  7. OCT
    18
    Rainer Keller
    Rainer Keller
    University of Stuttgart
    Tools for High-Performance Computing PDF
  8. OCT
    25
    Mark Dean
    Mark Dean
    EECS
    Technology Trends and Innovation Opportunities PDF

Upcoming Lunch Talks

  1. NOV
    1
    Azzam Haidar
    Azzam Haidar
    Unified Development for Mixed Multi-GPU and Multi-Coprocessor Environments using a Lightweight Runtime Environment PDF
  2. NOV
    8
    Jean-Charles Vasnier
    Jean-Charles Vasnier
    ORNL
    One OpenCL To Rule Them All PDF
  3. NOV
    15
    Jakub Kurzak
    Jakub Kurzak
    BEAST: A Transparent Autotuner for Accelerators PDF

Visitors

  1. Joseph Dobson
    Joseph Dobson from the Numerical Algorithms Group (NAG) will be visiting from October 21 through October 25. Joseph will be working with the PLASMA group.

Visitors

  1. Joseph Dobson
    Joseph Dobson from the Numerical Algorithms Group (NAG) will be visiting from October 21 through October 25. Joseph will be working with the PLASMA group.

Dates to Remember

SC13 Early Registration

Early registration for SC13 closes on October 15th! If you plan on going to SC and haven’t registered, please do so as soon as possible.