News and Announcements

CSE 11 Recap

A large contingent of current and past ICL researchers descended on Reno, Nevada in March for the SIAM Conference on Computational Science and Engineering (CSE11). The current team members who attended were Asim Yarkan, Azzam Haidar, Jakub Kurzak, Mathieu Faverge, Piotr Luszczek, and Stan Tomov; the ICL alumni attending were Hatem Ltaief, Emmanuel Aguillo, Marc Baboulin and Cedric Augonnet. Photographic evidence (a sample of which is seen above) suggests that, in addition to all the well received presentations they made, a good time was had by all.

Online registration for SAAHPC 2011 now open

The 2011 Symposium on Application Accelerators in High-Performance Computing will be held again in Knoxville on July 19th – 21st.

Online registration is now open for SAAHPC 2011. Through June 30, registration is discounted to $150; starting July 1, registration will be $175. Students can register for just $75. See http://saahpc.org to register.

The best paper on the use of GPUs for scientific computing presented at SAAHPC 2011 will be rewarded with an NVIDIA Tesla C2070 GPU computing processor, courtesy of NVIDIA. Submissions are due by May 6, 2011. See http://saahpc.ncsa.illinois.edu/submit/.

Supercomputers to increase capacity and speed

At the recent HPCC conference Jack gives a talk on Architecture-Aware Algorithms and Software for Peta & Exascale Computing. This talk is in 3 parts available on Youtube.
In other new, Jack explains the magnitude of the move from petascale to exascale in this Tennessee Journalist article, Supercomputers to increase capacity and speed.

Better computers allow better simulations and more confident predictions. Science today is built on three things – theory, experiments and computer simulations.

Recent Releases

PAPI CUDA Component Released

The CUDA component is a hardware performance counter measurement technology for the NVIDIA CUDA platform which provides access to the hardware counters inside the GPU. PAPI CUDA is based on CUPTI support – shipped with CUDA 4.0rc – in the NVIDIA driver library. In any environment where the CUPTI-enabled driver is installed, the PAPI CUDA component can provide detailed performance counter information regarding the execution of GPU kernels.
If users want to test the component, they will have to do a clean checkout from cvs. Also, we are planning on releasing PAPI 4.2 in a week or two in case you want to wait for an official release.

MAGMA 1.0 RC4 Released

MAGMA 1.0 RC4 is now available. This release includes the MAGMA sources! MAGMA 1.0 RC4 is intended for a single CUDA enabled NVIDIA GPU. It extends version 0.2 by adding support for the Fermi GPUs (see the sample performances for LU, QR, and Cholesky factorizations and LS solvers in complex arithmetic). For more details see the MAGMA 1.0 presentation.

Included are routines for the following algorithms:

  • LU, QR, and Cholesky factorizations in both real and complex arithmetic (single and double);
  • Hessenberg, bidiagonal, and tridiagonal reductions in both real and complex arithmetic (single and double);
  • Linear solvers based on LU, QR, and Cholesky in both real and complex arithmetic (single and double);
  • Eigen and singular value problem solvers in both real and complex arithmetic (single and double);
  • Mixed-precision iterative refinement solvers based on LU, QR, and Cholesky in both real and complex arithmetic;
  • MAGMA BLAS in real arithmetic (single and double), including gemm, gemv, symv, and trsm.

See the Software section for a download link.

MAGMA was also mentioned in the recent release of the PGI Insider, read the article here, Using GPU-enabled Math Libraries with PGI Fortran.

Interview

Charles Collins Then

Charles Collins

Where are you from, originally?

I was born in southern California, but raised in west Tennessee. My dad was in the Navy and retired when I was little. That’s when we moved.

Can you summarize your educational background?

After high school, I joined the Navy and worked as an aviation electronics technician. In 1994, I moved to Knoxville to attend UT, where I met my wife Julia. I eventually started working as a computer technician and left UT. I received an Information Technology BS from American Intercontinental University in 2004 and have been working as a Systems Administrator ever since.

Where did you work before joining ICL?

I spent most of the last few years working at URS corporation in Oak Ridge as a Systems Administrator. I managed Windows and Linux servers that provide web and database services to external customers.

Tell us how you first learned about ICL.

For years, I’ve had an interest in working with research-oriented organizations. I had been focusing on a career change to ORNL until I noticed there was an opening at ICL. I had not previously heard of ICL, but when I read the 2010 report, I became impressed with its role in the supercomputing industry. I immediately began pursuing the IT Administrator position there.

What made you want to work for ICL?

I have always been a fan of large computing environments. I regularly follow news concerning the Jaguar and Kraken systems and am very proud of east Tennessee’s role in this cutting-edge field. When I read about the bright people that work at ICL and the unique role that ICL plays in the supercomputing arena, I knew I wanted to be a part of that.

What are your interests/hobbies outside work?

I sail keelboats whenever I get the opportunity. I’ve been sailing on and off since 1992 and love it. I recently visited San Diego for Navy business and made sure to go sailing multiple times. I also have a passion for computer hardware and experimentation, so I regularly spend time working on computers at home.

Tell us something about yourself that might surprise people.

As a Navy reserve member, I was deployed to Iraq in 2005. While there, I served as a Seabee electrician and systems administrator at Al Taqaddam air base. I enjoyed working with the Iraqi people and hope to go back there as a tourist some day.

What will you be working on while at ICL?

I will be working with Paul Peltz to build and maintain servers and clusters as needed by the researchers.

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

I would love to work for NASA. I’m a big fan of their space exploration endeavors and would enjoy taking part in that. If they called me today and asked me to fly into space, I’d immediately say yes!

Recent Papers

  1. Haidar, A., H. Ltaeif, A. YarKhan, and J. Dongarra, Analysis of Dynamically Scheduled Tile Algorithms for Dense Linear Algebra on Multicore Architectures,” University of Tennessee Computer Science Technical Report, UT-CS-11-666, (also Lawn 243), March 2011.  (1.65 MB)
  2. Kurzak, J., S. Tomov, and J. Dongarra, Autotuning GEMMs for Fermi,” University of Tennessee Computer Science Technical Report, UT-CS-11-671, (also Lawn 245), April 2011.  (397.45 KB)
  3. Dongarra, J., M. Faverge, H. Ltaeif, and P. Luszczek, Exploiting Fine-Grain Parallelism in Recursive LU Factorization,” Proceedings of PARCO'11, no. ICL-UT-11-04, Gent, Belgium, April 2011.
  4. Tomov, S., MAGMA - LAPACK for GPUs , Atlanta, GA, Keeneland GPU Tutorial, April 2011.  (742.14 KB)
  5. Becker, D., M. Faverge, and J. Dongarra, Towards a Parallel Tile LDL Factorization for Multicore Architectures,” ICL Technical Report, no. ICL-UT-11-03, Seattle, WA, April 2011.  (425.45 KB)

Recent Lunch Talks

  1. MAR
    4
    Ala Rezmerita
    Ala Rezmerita
    University of Paris-Sud / INRIA Saclay
    Private Virtual Cluster PDF
  2. MAR
    11
    Marc Baboulin
    Marc Baboulin
    Universite Paris-Sud, parall team Project Grand-Large, INRIA Saclay
    Why and How to compute (cheaply) condition numbers PDF
  3. MAR
    18
    Azzam
    Azzam
    Parallel Reduction to Condensed Forms for Symmetric Eigenvalue Problems using Fine-Grained and Memory-Aware Kernels PDF
  4. MAR
    21
    Wesley Bland
    Wesley Bland
    A Fault Tolerant Runtime for Open MPI PDF
  5. MAR
    21
    Wesley Alvaro
    Wesley Alvaro
    ICL Linear Algebra Group Hat Tricks URL
  6. MAR
    25
    Yves Robert
    Yves Robert
    ENS Lyon et IUF
    Checkpointing strategies for parallel jobs PDF
  7. APR
    1
    Lynne Parker
    Lynne Parker
    CISML
    Overview of the Center for Intelligent Systems and Machine Learning PPTX
  8. APR
    8
    Heike
    Heike
    Methods for Heterogeneous Parallel Performance Measurement with GPUs PDF
  9. APR
    13
    Franz Franchetti
    Franz Franchetti
    CMU
    Spiral: Generating Efficient Programs for Emerging Parallel Platforms PDF
  10. APR
    15
    Piotr Luszczek
    Piotr Luszczek
    Three Little Modeling Problems PDF
  11. APR
    29
    Lizhe Wang
    Lizhe Wang
    Indiana University
    Cloud computing, a perspective computing

Upcoming Lunch Talks

  1. MAY
    6
    Mitch Horton
    Mitch Horton
    A Class of Hybrid LAPACK Algorithms for Multicore and GPU Architectures PDF
  2. MAY
    13
    Micah Beck
    Micah Beck
    EECS
    Memoization Gone Wild! PDF
  3. MAY
    20
    Peng Du
    Peng Du
  4. MAY
    27
    Jakub Kurzak
    Jakub Kurzak
    PULSAR PDF

People

  1. Franz Franchetti
    Franz Franchetti from CMU will be visiting us on April 13th. Franz has been involved in auto tuning and is one of the developers of Spiral project, http://www.spiral.net/.
  2. Eric Meek
    Past ICLer Eric Meek has re-joined the group and will be working as a consultant on the PAPI project.
  3. Vince Weaver
    Congratulations to Vince Weaver and his wife who gave birth to Kai Matthew Weaver on March 28th. He is 7lb 3ozs (3.26kgs) and both mother and baby are doing well.
  4. Sam Crawford
    Sam Crawford joined ICL on March 28th.  He will be taking on many of Scott Wells previous responsibilities.