News and Announcements
2020 Retreat
As with everything 2020, the annual ICL retreat was a bit different this year. Rather than hoof it over to Gatlinburg or Townsend, we all “dialed in” to the retreat from our own home offices, living rooms, and wherever else.
Zoom was a capable platform for two days of talks that covered student projects and summer internships; the lab’s progress in the areas of linear algebra, distributed computing, benchmarking, and performance analysis; and recaps of administrative procedures.
Serving as a kickoff to the fall semester, the 2020 retreat was a success by any measure. As we move through the rest of the year, ICL will continue to move forward and adapt where necessary. Here’s to next year’s retreat, where we can hopefully be more concerned about bear encounters than human malware.
Employment Opportunities at ICL
ICL is seeking full-time Research Scientists (MS or PhD) to participate in the design, development, and maintenance of numerical software libraries for solving linear algebra problems on large, distributed-memory machines with multi-core processors, hardware accelerators, and performance monitoring capabilities for new and advanced hardware and software technologies.
The prospective researcher will coauthor papers to document research findings, present the team’s work at conferences and workshops, and help lead students and other team members in their research endeavors in ongoing and future projects. Given the nature of the work, there will be opportunities for publication, travel, and high-profile professional networking and collaboration across academia, labs, and industry.
An MS or PhD in computer science, computational sciences, or math is preferred. Background in at least one of the following areas is also preferred: numerical linear algebra, HPC, performance monitoring, machine learning, or data analytics.
For more information check out ICL’s jobs page: http://www.icl.utk.edu/jobs.
Conference Reports
Best Paper at IEEE Cluster 2020
This year’s IEEE Cluster Conference went virtual in lieu of an on-the-ground meeting in Kobe, Japan. As in years past, IEEE Cluster 2020 served as a major international forum for presenting and sharing recent accomplishments and technological developments in the field of cluster computing as well as the use of cluster systems for scientific and commercial applications.
For ICL’s part, alum Xi Luo presented the DisCo team’s work on HAN: A Hierarchical AutotuNed Collective Communication Framework, which works towards addressing the increased scale and heterogeneity of new and upcoming HPC platforms and the challenges these trends present to the design of MPI libraries. With a task-based and modular design, HAN can easily swap out abstracted submodules while keeping tasks intact to adapt to new hardware. This strategy provides strong and flexible support for future HPC systems that also—owing to its autotuning component—significantly improves the default Open MPI implementation and achieves decent speedups against state-of-the-art MPI implementations in a battery of application tests.
Luo, X., W. Wu, G. Bosilca, Y. Pei, Q. Cao, T. Patinyasakdikul, D. Zhong, and J. Dongarra, “HAN: A Hierarchical AutotuNed Collective Communication Framework,” IEEE Cluster Conference, Kobe, Japan, IEEE Computer Society Press, September 2020.
This paper, which describes work undertaken for Xi’s PhD dissertation, also won a Best Paper Award at IEEE Cluster 2020. Congratulations to Xi and his coauthors!
EuroMPI 2020
This year’s EuroMPI was actually held in Austin, TX, though it wasn’t really there either. The virtual conference kicked off on September 21st and was Chaired by Wesley Bland—ICL alum and former DisCo Graduate Research Assistant.
Dong Zhong, current DisCo Graduate Research Assistant, presented his team’s work on “Using Advanced Vector Extensions AVX-512 for MPI Reduction.” You can watch his entire presentation above or read the paper here.
Smoky Mountains Computational Sciences and Engineering Conference
This year’s Smoky Mountains Computational Sciences and Engineering Conference (SMC 2020) was hosted virtually by Oak Ridge National Laboratory. No bear (or tiger) encounters this year, but the conference had a fully packed agenda and good attendance with over 350 people registered for the event.
For ICL’s part, Graduate Research Assistant Neil Lindquist presented his paper “Improving Performance of the GMRES Method Using Mixed Precision Techniques” on Wednesday, August 26th.
The GMRES method is used to solve sparse, non-symmetric systems of linear equations arising from many scientific applications. To reduce the amount of data movement and improve performance, Neil and his coauthors found that GMRES only needs double precision when computing the residual and updating the approximate solution and would still achieve double-precision accuracy with a mixed-precision approach in other parts of the solver.
SMC is always a popular workshop among ICLers, and we hope to be physically present again at SMC 2021.
Recent Releases
PLASMA 20.9.20 Released
PLASMA 20.9.20 is now available. The Parallel Linear Algebra Software for Multi-core Architectures (PLASMA) implements a set of fundamental linear algebra routines using the latest updates to the Open Multi-Processing (OpenMP) standard. PLASMA includes, among others, routines for solving linear systems of equations, linear least square problems, parallel BLAS, and parallel matrix norms.
New changes for PLASMA 20.9.20 are listed below.
Features:
- added support for detecting Apple’s Accelerate Framework as BLAS and LAPACK implementations;
- added GPU offload option detection with the MAGMA library;
- made PLASMA pure OpenMP by removing code that relies on POSIX threads; and
- added version reporting in PLASMA tester.
Bug fixes:
- fixed support for wider range of CMake versions and
- fixed Fortran examples’ syntax.
Check out PLASMA’s BitBucket repository to download the tarball.
Interview
Maksim Melnichenko
Where are you from, originally?
I am originally from Kaliningrad, Russia.
Can you summarize your educational background?
I graduated from Millersville University of Pennsylvania in May 2020 with two Bachelor of Science degrees: (1) Applied Mathematics and (1) Computer Science. Currently, I am pursuing an early stage PhD program in Mathematics at UTK.
Where did you work before joining ICL?
My past experience is rather modest—I worked on an undergraduate research project during my senior year and have been working for the Department of Mathematics as a tutor since fall 2018 and up until graduation.
How did you first hear about the lab, and what made you want to work here?
Throughout my undergraduate research, I have seen many references to the ICL’s papers.
Additionally, I heard many things about the lab from my undergraduate advisor. I visited the lab during Fall break in 2019, talked to Dr. Luszczek, and solidified my decision to apply to ICL.
What is your focus here at ICL? What are you working on?
I work as a Graduate Research Assistant in the Linear Algebra group. So far, my work is related to overviewing and testing the singular value decomposition (SVD)–related routines of PLASMA. As for potential future work, I am hoping to continue the research in optimizing and enhancing SVD algorithms.
What are your interests/hobbies outside of work?
My primary hobbies are graphical artwork and tattooing. I am also interested in historical and philosophical literature, the video game industry and design, and playing guitar.
Tell us something about yourself that might surprise people.
The location of my hometown tends to surprise people that are not especially interested in geography.
If you weren’t working at ICL, where would you like to be working and why?
Most realistically, I would take a TA position in the Department of Mathematics at UTK, as I would still be pursuing a PhD in Mathematics.
