The Utpal Banerjee Distinguished Lecture in High Performance Computing

Featuring: David J. Kuck, Intel Fellow and Professor Emeritus

Title: Optimizing a Computation

Abstract: Optimizing a computation presents a problem with two halves: optimizing SW for a given HW system, and optimizing HW for a given SW structure (program + data). This is difficult for the obvious reasons that HW and SW keep changing, but more deeply, because optimality is a function of many variables. We will review 50 years of studying this still unsolved problem, much of which has been done at the University of Illinois Urbana-Champaign (UIUC). Crucial issues are to understand which variables matter for a given computation, and how to test for optimality in reasonable time. We will discuss emerging, practical methods to solve the problem of optimizing SW for a given HW system, relative to a given set of program transformations.

Biography: David J. Kuck is an Intel Fellow working on HW/SW codesign in Intel’s Software and Advanced Technology Group. He was a Professor of CS/ECE at UIUC and founder of the CSRD. He was a founder and Chairman of KAI from 1979 until 2000 when it was acquired by Intel. He is a Fellow of the IEEE, ACM and AAAS, and has received the IEEE Piore Award, the IEEE Computer Society’s Babbage Award and Computer Pioneer Award, the ACM-IEEE Eckert-Mauchly and Kennedy Awards, the 2018 Okawa Prize, and is a member of the National Academy of Engineering.

Biography of Utpal Banerjee

An adjunct professor of Computer Science at the University of California, Irvine, Utpal Banerjee spent nearly 20 years of his career at Intel in the Software Solutions Group, where he developed techniques to enhance the performance of multi-core processors.

In the early days of parallel computing while working as a graduate student for Illinois CS Professor David Kuck, Banerjee developed a strategy for automatically analyzing a loop and determining whether it could be executed in parallel. His method, later dubbed Banerjee’s test, was a simple, fast, and effective data dependence test that has been widely used for compiler development and has had a big impact on advancing compiler technology.

While at Intel, he developed a formal representation and generalization of loop transformation techniques. By formalizing the transformation of code, Banerjee created the foundations of modern work on program restructuring, which is now known as the polyhedron model. In the 1990s, he wrote several books on loop dependence and transformations for restructuring compilers. Banerjee is a fellow of ACM and IEEE.

About The Utpal Banerjee Distinguished Lecture in High Performance Computing

When Utpal Banerjee died in 2017, members of his family said they were only beginning to understand the impact the Illinois Computer Science alumnus had on high performance computing. As a result, his daughter, Sanchita Banerjee Saxena, and her family have established the Utpal Banerjee Distinguished Lecture Series in High Performance Computing. Through their gift they are able to both honor his memory and provide a platform for the department to showcase one of the areas in which it excels, and to remember one of its own in a lasting way.

“It was really important for us to honor his memory at the University of Illinois in some way. The lecture series was something we knew he would be interested in,” said Saxena. “My dad always wanted to engage with new ideas that were coming up, he always loved interacting with both junior and senior scholars and taking part in interesting conversations.”