Virginia Tech Professor of Computer Science and John W. Hancock Professor of Engineering Dimitrios Nikolopoulos has been honored with the Best Paper award distinction at the 2020 Design, Automation and Test in Europe Conference. 

The paper, “DEFCON: Generating and Detecting Failure-Prone Instruction Sequences via Stochastic Search,” represents one of the first systematic efforts in identifying failure-prone code within any application that can potentially threaten the correct functionality of any system under adverse operating conditions. It was selected from among 748 submissions authored by 3,107 researchers across 45 countries.

The DATE conference is the world's top-tier venue on electronic design automation, bringing together thousands of experts from academia and industry on hardware and software design, security, and testing and manufacturing of electronic circuits and systems. The conference is sponsored by the Institute for Electrical and Electronic Engineers and the Association for Computing Machinery

Nikolopoulos collaborated with Ioannis Tsiokanos, a Ph.D. candidate, and Georgios Karakonstantis, an associate professor of electronics, electrical engineering and computer science, both from Queen’s University Belfast, as well as Giorgis Georgakoudis, a computer scientist at the Lawrence Livermore National Laboratory.

According to the paper, modern processors exhibit an alarmingly increasing rate of hardware errors due to manufacturing variability. Nikolopoulos shares that the research community has speculated for many years that these errors are triggered by specific instruction patterns in software. 

“Finding these patterns would lead to better error prevention or mitigation methods, including methods that would allow the hardware to operate with extended performance and energy-efficiency margins,” said Nikolopoulos.

The awarded paper is the first to identify with very high probability the critical, error-inducing instruction sequences in programs. The work exploits the error-inducing instruction sequences in several methods, including clock stretching, a method that improves the energy-efficiency of a pipelined processor by more than 25 percent, and precise fault injection for accurately capturing the resilience of software to errors. 

Throughout his career, Nikolopoulos has received nine Best Paper awards. His earlier paper awards recognized excellence and significant research advances in memory management, heterogeneous parallel programming, runtime adaptation of programs to dynamic execution environments, and more recently, edge computing architectures for smart manufacturing.

Through fundamental scholarship on computer systems, Nikolopoulos has made contributions to the global computing systems research community. He has published 59 peer-reviewed journal articles and 132 peer-reviewed papers in highly regarded archival conference proceedings. The total value of his research awards is $101.7 million and his own share of these awards is $38.9 million.

He received his bachelor’s degree, master’s degree, and Ph.D. from the University of Patras.

— Written by Jenise L. Jacques