David J. Kuck

From ETHW

David J. Kuck
David J. Kuck

Biography

David J. Kuck is an inspirational leader in the development of tools for high-performance computing. Every computer now employs parallel processing to deliver the performance needed for applications from scientific discovery to machine learning. Kuck is a pioneer in this very important field as he invented the concept of parallelizing compilers, which can automatically transform sequential programs to run on parallel machines. The work he spearheaded on dependence-based restructuring compiler technology is one of the most significant contributions to the use of programming languages for parallel computing. This work, started more than 50 years ago, created a completely new line of research that has produced an immense number of publications that continue to this day. In particular, Frances Allen’s PTRAN restructuring compiler and Ken Kennedy’s Rn and PFC restructuring compilers were inspired by Kuck’s Parafrase, as were vectorizing compilers built by IBM, CRAY, Hitachi and NEC. In 1979, Kuck started his own firm, Kuck and Associates Inc. (KAI), which licensed its technology to all major supercomputer vendors at that time. Intel acquired KAI in 2000. Today, all state-of-the-art compilers, including llvm, gcc, Intel’s icc, and IBM’s xlc implement auto-vectorization, use the approach invented by Kuck and his students. Kuck’s contributions to programming language and benchmarking are equally influential. The twenty-plus-year-old OpenMP is not only one of the most widely used language extensions for shared-memory parallelism, but is also an important inspiration for the development of new parallel programming languages and extensions.

An IEEE Life Fellow, Kuck was an Intel Fellow (2000 to 2023), and is Professor Emeritus, University of Illinois at Urbana-Champaign, Champaign, Illinois, USA. Kuck is the recipient of the 2024 IEEE Frances E. Allen Medal for "pioneering work in vector and parallel computer architecture, software, and compilers that enables many performance-sensitive applications".

Further Reading