2026 ICL Winter Reception
ICL hosted its 2026 Winter Reception on March 26 at Chesapeake’s, bringing together current and former ICL members for an evening of great food and conversation. Thanks to the ICL operations team for coordinating the event.
ICL 2025/26 Report
The latest edition of the ICL annual report is now available to download as a PDF and print copies are available from ICL’s main office.
2026 Sparse BLAS Workshop
The Sparse BLAS Workshop 2026, held March 23–26 at the University of Tennessee, Knoxville, continued a multi-year effort to define and advance a standardized interface for sparse linear algebra. Organized by Hartwig Anzt, the workshop brought together participants from academia, national laboratories, and industry to refine proposed APIs, evaluate reference implementations, and discuss practical pathways for adoption across emerging architectures. Special thanks to the ICL operations team for their support in making the event a success.
Scientific Computing in the Age of AI
A recent perspective co-authored by Jack Dongarra, Dennis Gannon, and Daniel Reed offers a forward-looking view of how scientific computing must evolve in response to the rapid rise of artificial intelligence. Titled “Ride the Wave, Build the Future: Scientific Computing in an AI World,” the piece serves as a strategic manifesto for the field at a moment of significant transition.
At its core, the paper argues that the “center of gravity” in high-performance computing has shifted toward hyperscale AI platforms. With industry investment and innovation now largely driven by generative AI, the scientific computing community must adapt by leveraging AI-driven advances while continuing to build systems tailored to scientific discovery. The authors frame this shift through a central idea: scientific computing must both “ride the wave” of AI by taking advantage of emerging hardware and software ecosystems, and “build the future” through targeted innovation in next-generation systems.
The piece has sparked broader discussion across the community. Co-author Daniel Reed expanded on these ideas in a follow-up blog post, “HPC In An AI World,” while additional perspective can be found in community coverage such as the HPCwire feature, “HPC Is Riding AI’s Coattails. So Now What?”
MAGMA Team Set to Receive New Funding for GPU Research Collaboration
ICL has received new funding to advance the MAGMA project through a new collaboration with Lawrence Livermore National Laboratory. Led by Ahmad Abdelfattah and Natalie Beams, the overall goal of this effort is to develop improvements in the MAGMA library that target large-scale GPU simulations at LLNL, including the MARBL code and the MFEM library. Planned work includes exploring batch linear algebra and FP64 emulation for emerging hardware.
Conference Reports
KAUST Research Conference on Mathematical and Data Sciences
In January, Jack Dongarra visited King Abdullah University of Science and Technology to speak at the KAUST Research Conference on Mathematical and Data Science (MDS26), which brought together international experts to explore advances in modeling, data, and simulation. During the visit, he connected with colleagues including Bill Gropp (University of Illinois Urbana-Champaign), David Keyes (KAUST), and ICL alum Rabab Al-Omairy (Massachusetts Institute of Technology).
Dagstuhl Seminar Advances Research in Reduced and Mixed Precision Computing
Leibniz Center for Informatics Dagstuhl Seminar 26081 on Reduced and Mixed Precision Computing for Science and Engineering Applications brought together leading researchers Feb. 15–20, 2026 in Dagstuhl to examine how reduced-precision arithmetic is reshaping numerical algorithms and performance on modern computing architectures. This seminar explored key challenges including balancing efficiency and accuracy, developing new algorithmic and data structure approaches, and ensuring reliable scientific results through rigorous analysis. Discussions also highlighted the growing influence of AI-driven workloads and emerging hardware in accelerating the adoption of mixed-precision methods, as well as the need for improved programming models, benchmarking strategies, and interdisciplinary collaboration. A strong group of current and former ICL researchers participated, including Emmanuel Agullo, Hartwig Anzt, Alfredo Buttari, Jack Dongarra, Julien Langou, Hatem Ltaief, Piotr Luszczek, and Theo Mary.
Recent Releases
MAGMA 2.10.0
The MAGMA development team is pleased to announce MAGMA-2.10.0.
MAGMA (Matrix Algebra on GPU and Multi-core Architectures) is a high performance library for dense linear algebra algorithms using GPUs.
Release Highlights:
– New functionality:
• Batch SVD and batch QR + SVD
• Batch UNMQR/ORMQR
– Non-uniform batch LU factorization without pivoting
– Performance improvements:
• Batch Cholesky factorization
• Batch triangular solve
– Other bug fixes and improvements
For more details, please visit: https://icl.utk.edu/magma
Interview
Keita Teranishi
Tell us a little about your background and what years you were at ICL.
I became part of ICL during the winter of 1997 while I was a junior undergraduate CS student at UTK, and worked until the end of 1998. At the time, I was searching for an on-campus job because it is only an option for foreign undergraduate students. I reached out to Ms. Mayo, the undergraduate advisor, who kindly introduced me to Jack and guided me toward joining the group.
What prompted you to join ICL back then and what did you work on during your time at ICL?
During the spring semester of 1998, I focused on developing the C-Interface to BLAS (CBLAS). Under the supervision of Clint Whaley, who was concurrently working on Atlas, I completed the entire CBLAS implementation by the end of the semester, which was my final semester in UTK Undergraduate Program. Afterward, I transitioned into a Graduate Research Assistant role, where I contributed to projects involving ScaLAPACK and LAPACK. My work in this position encompassed C-interface development, benchmarking, and tester creation, allowing me to further deepen my expertise in high-performance computing software.
Where are you working and what are you working on currently?
At Oak Ridge National Laboratory, I am leading the Programming Systems Group, where I lead three major projects in collaboration with multiple institutions. These initiatives, all supported by the DOE Advanced Scientific Computing Program, focus on advancing software ecosystems for parallel programming models, developing high-productivity programming languages tailored for high-performance computing, and innovating in AI-driven code synthesis and testing for HPC environments.
In what ways did working at ICL prepare you for what you do now, if at all?
My experience at ICL provided me with a comprehensive understanding of high-performance computing research and its practical applications. This broad perspective has significantly benefited my career, with valuable insights that have proven essential in both industry and academic settings.
What are some of your favorite memories from your days in the group?
Friday Lunch!
What are some of your interests/hobbies outside of work?
I have a deep passion for board games and currently own a collection of over 300 titles. My ultimate goal is to continue expanding this library and eventually experience playing each and every one of them.
