Predicting the Impact of Functional Approximation: from Component- to Application-Level

Marcello Traiola (1) , Alessandro Savino (2) , Mario Barbareschi (3) , Stefano Di Carlo (2) , Alberto Bosio (1)

1 TEST - Test and dEpendability of microelectronic integrated SysTems
2 Polito - Politecnico di Torino = Polytechnic of Turin
3 University of Naples Federico II = Università degli studi di Napoli Federico II

Approximate Computing (AxC) trades off between the level of accuracy required by the user and the actual precision provided by the computing system to achieve several optimizations such as performance improvement, energy and area reduction etc. Several AxC techniques have been proposed so far in the literature. They work at different abstraction levels and propose both hardware and software implementations. The common issue of all existing approaches is the lack of a methodology to estimate the impact of a given AxC technique on the application-level accuracy. In this paper we propose a probabilistic approach to predict the relation between component-level functional approximation and application-level accuracy. Experimental results on a set of benchmark applications show that the proposed approach is able to estimate the approximation error with good accuracy and very low computation time.

Mots clés: Approximate computing; Functional approximation; Quality Metrics; Bayesian Networks

Domaines: Micro et nanotechnologies/Microélectronique; Informatique [cs]

Marcello Traiola, Alessandro Savino, Mario Barbareschi, Stefano Di Carlo, Alberto Bosio. Predicting the Impact of Functional Approximation: from Component- to Application-Level. 24th International Symposium on On-Line Testing And Robust System Design (IOLTS), Jul 2018, Platja d'Aro, Spain. pp.61-64, ⟨10.1109/IOLTS.2018.8474072⟩. ⟨hal-03094581⟩