Nithya Narayanamurthy received an award for her paper describing a unique way of reducing computing errors at the European Dependable Computing Conference this week. Narayanamurthy developed her research while completing her Masters in Applied Science degree in Professor Pattabiraman’s laboratory in 2015, and is currently working at Oracle Labs.
Typically, application developers use compiler optimizations to speed up their programs. Unfortunately, traditional compiler optimizations are believed to make programs more vulnerable to errors, thus decreasing their error resilience. As a result, developers are faced with a dilemma to choose between performance and reliability. Narayanamurthy’s paper however finds that by judiciously choosing the optimizations, it is possible to make programs both faster and retain their error resilience. This is the proverbial equivalent of having one’s cake and eating it too.
In the award-winning paper, Narayanamurthy and her co-authors Profs. Pattabiraman and Ripeanu use an algorithm inspired by natural selection, a genetic algorithm to find optimization sequences. The genetic algorithm tests an initial set of solutions and crosses over two good options or mutates one good option to produce a second generation of better solutions. You can find out more about this research in their paper Finding Resilience-friendly Compiler Optimizations Using Meta-heuristic Search Techniques.
The European Dependable Computing Conference is a leading European venue for presenting and discussing the latest research, industrial practice and innovations in dependable and secure computing. This is the second time a student from Prof. Pattabiraman’s group has won a best paper award in this conference.
Professor Pattabiraman’s approach to overcoming inevitable computer hardware faults is to leave the hardware alone and use software to fix the problem instead. Prof. Pattabiraman uses algorithms to maximize efficiencies in energy use as well as to increase the resilience of the system. The above paper is an illustration of this approach and demonstrates its power.
Image: Nithya Narayanamurthy (top left) sightseeing with research colleagues