News and Announcements

Welcome to ICL

ICL-claxton

As we get underway for the 2014 – 2015 academic year, we would like to welcome some new students to the Innovative Computing Laboratory. Every year, ICL recruits some of the best and brightest in Computer Science and Computer Engineering to help push ICL and UTK to the forefront of High Performance Computing research.

The Distributed Computing group (DisCo) picked up 4 students this year—David Eberius (GRA), Xi Luo (GRA), Stephen Richmond (URA), and Chunyan Tang (GRA). The Performance Analysis group also picked up a new student this fall—Sangamesh Ragate (GRA). Welcome to ICL!

New ICL Website

Under the creative direction of David Rogers, ICL’s communications team has been developing a new website for ICL. The new site utilizes a Drupal back end and features a completely new look and advanced publications database. David gave a live demo of the site during the ICL retreat, and now invites ICLers to check out the website for themselves. Please keep in mind that the website is currently in Alpha, so there are some additional features that will be rolled out for the final release, and the up-time is not guaranteed.

URL: http://www.icl.utk.edu/

website_screenshot

Conference Reports

2014 ICL Retreat

The 2014 ICL Retreat brought the team back to the RT Lodge in Maryville, TN. The best weather in recent memory served as a backdrop for the ICLers, new and old, while the crew at the RT lodge took very good care of their guests. On the business end, the retreat was two solid days of talks covering student projects and summer internships, the lab’s progress in the areas of linear algebra, distributed computing, benchmarking, and performance analysis, along with recaps of administrative procedures.

It wasn’t all so serious, and the RT Lodge catered great food and plenty of refreshments for ICLers who seemed more thirsty than hungry. Serving as a kickoff to the fall semester, the retreat encourages us to keep our focus sharp, our determination strong, and to have another great year at ICL!

Recent Releases

MAGMA 1.5.0 Released

MAGMA 1.5.0 is now available. This release provides performance improvements for SVD and eigenvector routines and adds sparse routines. More information is given in the MAGMA: a New Generation of Linear Algebra Libraries for GPU and Multicore Architectures presentation as well as the MAGMA Quick Reference Guide. The MAGMA 1.5.0 release adds the following new functionalities:

  • SVD using Divide and Conquer (gesdd);
  • Nonsymmetric eigenvector computation is multi-threaded (trevc3_mt);
  • Sparse functions; and
  • Doxygen documentation.

Parameters (trans, uplo, etc.) now use symbolic constants (MagmaNoTrans, MagmaLower, etc.) instead of characters (‘N’, ‘L’, etc.). Converters are provided to translate these to LAPACK, CUBLAS, and CBLAS constants. There are also improved testers and numerous bug fixes.

Visit the software page to download the tarball.

Interview

Reazul Hoque Then

Reazul Hoque

Where are you from, originally?

I am from Bangladesh. It is a small country with high (read: extremely high) population density, sandwiched between India and Myanmar. I am from a city named Chittagong. I was born and raised there until I moved to the capital (Dhaka) for my undergraduate studies.

Can you summarize your educational background?

I completed my undergraduate studies in Computer Science and Information Technology at IUT (Islamic University of Technology). Currently, I am a PhD student here at UT and working at ICL as a Graduate Research Assistant.

Tell us how you first learned about ICL.

When searching for a graduate school, I applied to 10 different universities in the US. Funding was mandatory for me to pursue my higher studies. I was interested in distributed computing, and I first learned about ICL when I was applying to UT.

What made you want to work for ICL?

ICL is one of the best research labs in the world for HPC. Pursuing a PhD is important to me, and I always wanted to work in a place with a cutting edge research agenda, awesome mentors, and rich resources—both in personnel and infrastructure. For me, when considering all of those factors, ICL was the best match.

What are you working on while at ICL?

I am working on an interface for PaRSEC under the guidance of Geroge Bosilca, Anthony Danalis, Thomas Herault, and Aurelien Bouteiller. This new interface will allow users to insert tasks in the PaRSEC scheduler more easily and dynamically. This new interface also has a different approach to building DAGs and representing dependencies than the current interface.

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

I would like to work for a software company like Linkedin, MathWorks, etc. if I wasn’t working for ICL. The learning opportunities in these sorts of companies are tremendous and the working environments are great too.

What are your interests/hobbies outside work?

I like sports a lot and I used to play accordingly (a lot) too. I like playing Cricket, Tennis, Soccer, pool/billiards, table tennis, swimming. I also like shooting (guns of course).

I am also learning photography and one of my favorite leisure time activities is driving around in the beautiful city of Knoxville. The fact that I love driving is implied here. 🙂

Tell us something about yourself that might surprise people.

Since my sophomore year I have been a big fan of web development, and I love making web applications. The fact that the front end is slick and can be coded pretty easily is one of the major reasons behind my fascination with it. Also, the result of the development can be seen quickly, which is a bonus.

Recent Papers

  1. Genet, D., A. Guermouche, and G. Bosilca, Assembly Operations for Multicore Architectures using Task-Based Runtime Systems,” Euro-Par 2014, Porto, Portugal, Springer International Publishing, August 2014.  (481.52 KB)
  2. Baboulin, M., J. Dongarra, and R. Lacroix, Computing Least Squares Condition Numbers on Hybrid Multicore/GPU Systems,” International Interdisciplinary Conference on Applied Mathematics, Modeling and Computational Science (AMMCS), Waterloo, Ontario, CA, August 2014.  (130.18 KB)
  3. Dong, T., A. Haidar, P. Luszczek, J. Harris, S. Tomov, and J. Dongarra, LU Factorization of Small Matrices: Accelerating Batched DGETRF on the GPU,” 16th IEEE International Conference on High Performance Computing and Communications (HPCC), Paris, France, IEEE, August 2014.  (684.73 KB)
  4. Boillot, L., G. Bosilca, E. Agullo, and H. Calandra, Task-Based Programming for Seismic Imaging: Preliminary Results,” 2014 IEEE International Conference on High Performance Computing and Communications (HPCC), Paris, France, IEEE, August 2014.  (625.86 KB)
  5. Dong, T., A. Haidar, S. Tomov, and J. Dongarra, A Fast Batched Cholesky Factorization on a GPU,” International Conference on Parallel Processing (ICPP-2014), Minneapolis, MN, September 2014.  (1.37 MB)
  6. McCraw, H., J. Ralph, A. Danalis, and J. Dongarra, Power Monitoring with PAPI for Extreme Scale Architectures and Dataflow-based Programming Models,” 2014 IEEE International Conference on Cluster Computing, no. ICL-UT-14-04, Madrid, Spain, IEEE, September 2014. DOI: 10.1109/CLUSTER.2014.6968672  (3.45 MB)
  7. Haugen, B., and J. Kurzak, Search Space Pruning Constraints Visualization,” VISSOFT'14: 2nd IEEE Working Conference on Software Visualization, Victoria, BC, Canada, IEEE, September 2014.  (1.32 MB)
  8. Aliaga, J. I., H. Anzt, M. Castillo, J. C. Fernández, G. León, J. Pérez, and E. S. Quintana-Orti, Unveiling the Performance-energy Trade-off in Iterative Linear System Solvers for Multithreaded Processors,” Concurrency and Computation: Practice and Experience, vol. 27, issue 4, pp. 885-904, September 2014. DOI: 10.1002/cpe.3341  (1.83 MB)
  9. McCraw, H., A. Danalis, G. Bosilca, J. Dongarra, K. Kowalski, and T. Windus, Utilizing Dataflow-based Execution for Coupled Cluster Methods,” 2014 IEEE International Conference on Cluster Computing, no. ICL-UT-14-02, Madrid, Spain, IEEE, September 2014.  (260.23 KB)

Recent Conferences

  1. SEP
    RaPyDLI Working Meeting Palo Alto, CA, California
    Jakub Kurzak
    Jakub
    Piotr Luszczek
    Piotr
    Terry Moore
    Terry
    Jakub Kurzak, Piotr Luszczek, Terry Moore
  2. SEP
    Resilience Building Blocks Proposal Meeting Argonne, Illinois
    Aurelien Bouteiller
    Aurelien
    George Bosilca
    George
    Thomas Herault
    Thomas
    Aurelien Bouteiller, George Bosilca, Thomas Herault

Upcoming Conferences

  1. OCT
    Aurelien Bouteiller
    Aurelien
    Aurelien Bouteiller
  2. OCT
    Super Fall Meeting College Park, Maryland
    George Bosilca
    George
    George Bosilca
  3. OCT
    IEEE International Conference on Big Data Washington, District of Columbia
    Ichitaro Yamazaki
    Ichitaro
    Ichitaro Yamazaki

Recent Lunch Talks

  1. AUG
    22
    Tracy Rafferty
    Tracy Rafferty
    Conference travel PDF
  2. AUG
    29
    Gregoire Pichon
    Gregoire Pichon
    INRIA
    Divide and Conquer: a symmetric tridiagonal eigensolver in PLASMA PDF
  3. SEP
    5
    Theo Mary
    Theo Mary
    INP-ENSEEIHT
    Performance Study of a Randomized Low-rank Approximation using multi-GPU PDF
  4. SEP
    12
    George Ostrouchov
    George Ostrouchov
    ORNL
    Taking R to Big Platforms and Supercomputers with pbdR
  5. SEP
    19
    Simplice Donfack
    Simplice Donfack
    Improve the applicability of highly efficient stencil compilers to a wider class of problems PDF
  6. SEP
    26
    Azzam Haidar
    Azzam Haidar
    Towards Batched Linear Solvers on Accelerated Hardware Platforms PDF

Upcoming Lunch Talks

  1. OCT
    3
    Hartwig Anzt
    Hartwig Anzt
    Asynchronous Iterative Algorithm for Computing Incomplete Factorizations on GPUs PDF
  2. OCT
    10
    Alfredo Buttari
    Alfredo Buttari
    ENSEEIHT
    Improving multifrontal solvers by means of Block Low-Rank approximations PDF
  3. OCT
    17
    Florent Lopez
    Florent Lopez
    ENSEEIHT
    Sparse direct solvers on top of runtime systems PDF
  4. OCT
    24
    Yves Robert
    Yves Robert
    Assessing general-purpose algorithms to cope with fail-stop and silent errors PDF
  5. OCT
    31
    Aurelien Bouteiller
    Aurelien Bouteiller
    UCCS: A Communication Substrate for Open SHMEM (and more) PDF

People

  1. James Ralph
    James Ralph accepted a job offer with SalesForce.com, where he will be working in their new performance division in San Francisco, CA. James departs the lab at the end of September. Congratulations and good luck, James!

Dates to Remember

SC14 Early Registration

The SC14 early registration deadline is October 15th. Registering before this date can save ICL up to $275/person, so please take advantage of early registration!

If you are a SIGHPC member, you maybe be eligible for a further discount when registering for SC14. See the SC14 registration form for details.