The Georgia Institute of Technology’s School of Computational Science and Engineering (CSE) recently hosted well-known computer scientist and mathematician James Demmel. 

Demmel, a distinguished professor at the University of California, Berkeley (UC Berkeley), presented a lecture titled, Communication-Avoiding Algorithms for Linear Algebra and Beyond. The presentation discussed his team’s efforts to make algorithms more self-regulating while reducing energy loss and the time associated with moving data between memory levels or amongst processors. During the lecture, Demmel shared findings that showed the substantial speed increase gained by algorithms with reduced communication.

The lecture also doubled as a reunion for Demmel. Demmel has two former students, Rich Vuduc and Jason Riedy, working in the College of Computing. He advised them both during their Ph.D. program at UC Berkeley. The three still collaborate on many research projects dealing with high-performance computing and parallel algorithms.

A noted authority in computer science, Demmel examines topics such as numerical linear algebra, high-performance computing, and communication avoiding algorithms. He is known for his work on the Linear Algebra Package, better known as LAPACK, a software library for handling the implementation of high-performance computing. During the 1999 Supercomputing conference, Demmel, along with his collaborators, won the Carl Benz Award for writing Prometheus, an algorithm that concurrently resolves multiple problems in engineering and mathematical physics.

Demmel is a fellow in various scientific organizations such as the Association for Computing Machinery, American Mathematical Society, the Institute of Electrical and Electronics Engineers (IEEE), and the Society for Industrial and Applied Mathematics, to name a few. Furthermore, Demmel was one of two scientists honored with the Leslie Fox Prize for Numerical Analysis in 1986. In 1993, Demmel won the J.H. Wilkinson Prize in Numerical Analysis and Scientific Computing, and in 2010, he won the IEEE's Sidney Fernbach Award "for computational science leadership in creating adaptive, innovative, high-performance linear algebra software."

Demmel gave the first of two distinguished lectures organized by CSE this semester. The CSE Distinguished Lecture Series are quarterly talks from highly regarded experts on a variety of computational topics. The next lecture will feature Adolfy Hoisie, physical and computational sciences manager for the Pacific Northwest National Laboratory, on March 29 at 11 a.m. in the Klaus Advanced Computing Building, Room 1116.