Age of Mixed-Precision: Algorithms, Libraries, and Applications

Organizers: Piotr Luszczek (MIT LLSC & UTK ICL) and Kurt Keville (Massachusetts Institute of Technology)

This is a special special session of IEEE HPEC 2026.

AI has triggered an explosion of new capabilities for computing with mixed-precision number formats. Algorithms that can take advantage of these innovations can be accelerated by many orders of magnitude. Future progress in computational methods is dependent on the development of these new approaches. This session will focus on the use of mixed-precision formats across a variety of hardware platforms. The session will highlight advancements in algorithms, software, libraries, and applications that have incorporated these new number formats and leveraged their potential for improvements in performance, bandwidth, and energy consumption.

Invited talks:

11:00-11:25 Recent Progress in Ozaki Schemes by Katsuhisa Ozaki, Shibaura Institute of Technology
11:25-11:50 Variable Precision Formats that Preserve Dynamic Range by John Gustafson, Arizona State Universty
11:50-12:15 Mixed model, mixed method, and mixed precision Runge–Kutta method by Sigal Gottlieb, UMass Dartmouth

Abstracts

Recent Progress in Ozaki Schemes: The Ozaki schemes emulate high-precision arithmetic by leveraging low-precision arithmetic, whose performance has been improving rapidly, driven by the growing demand for AI. In this talk, I will first present the current state of the art, followed by the remaining challenges and some approaches to overcoming them. I will then introduce several recent technologies that are attracting considerable attention.

Bios

Prof. Katsuhisa Ozaki: is a Professor in the Mathematical Sciences Program, Faculty of Systems Engineering and Science, Shibaura Institute of Technology, Japan. He received his Ph.D. in Engineering from Waseda University in 2007. His research interests include numerical linear algebra, reliable computing, high-accuracy numerical algorithms, and mixed-precision computing.
He is best known for the development of the Ozaki scheme, a family of techniques for emulating high-precision matrix multiplication using lower-precision, high-throughput arithmetic. His recent work focuses on exploiting low-precision matrix engines on modern accelerators while retaining numerical accuracy and reproducibility. He has also worked extensively on verified numerical computation and rounding-error analysis.
Prof. John L. Gustafson: (www.johngustafson.net) is Chief Scientist of Vq Research and a Visiting Scholar at Arizona State University. He is the inventor of several novel forms of computer arithmetic first introduced in his 2015 book, The End of Error: Unum Computing. He is best known for his 1988 argument showing that parallel processing performance need not be limited by "Amdahl's law," now generally known as Gustafson's law. Previously, he has been Senior Fellow and Chief Product Architect at AMD and a Director of Intel Labs. He is a recipient of the inaugural Gordon Bell Prize and is a Golden Core member of IEEE.