Dr. Mooney’s paper entitled “Classification-based Improvement of Application Robustness and Quality of Service in Probabilistic Computer Systems” is based on collaboration among the coauthors who represent TU Dortmund (Germany), Nanyang Technical University (Singapore), Georgia Tech, and Rice University. Future semiconductors no longer guarantee permanent deterministic operation and are expected to show probabilistic behavior due to lowered voltages and shrinking structures. However, instructions handling control flow or pointers still require determinism. This paper is the first to identify these instructions and apply a transient error classification to probabilistic circuits considering voltage distributions, ensuring that probabilistic effects only affect unreliable operations that accept a certain level of impreciseness. To evaluate, this team analyzed the robustness and quality-of-service of an H.264 video decoder and mapped unreliable arithmetic operations onto probabilistic components of an ARM-based simulator, while remaining operations used deterministic components.

Dr. Mooney’s coauthors on this paper are Andreas Heinig (TUD), Florian Schmoll (TUD), Peter Marwedel (TUD), Krishna Palem (Rice and NTU), and Michael Engel (TUD). Dr. Mooney took part in this work while he was a visiting professor at the Institute of Sustainable and Applied Infodynamics at NTU Singapore.