Error Analysis for Loop Programs using Anytime Instructions in Approximate Computing

Conference: MBMV 2020 – Methoden und Beschreibungssprachen zur Modellierung und Verifikation von Schaltungen und Systemen - GMM/ITG/GI-Workshop
03/19/2020 - 03/20/2020 at Stuttgart, Deutschland

Proceedings: GMM-Fb. 96: MBMV 2020

Pages: 4Language: englishTyp: PDF

Keszocze, Oliver; Koenig, Moritz; Brand, Marcel; Teich, Juergen (Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Erlangen, Germany)

Approximate Computing is a novel design paradigm sacrificing computational accuracy for gains in other non-functional properties. We present a tool that allows to automatically analyze approximated loop programs for the resulting error and range of output values. Our tool allows to model non-uniform distributions for the input variables. We further support the novel concept of Anytime Instructions: Anytime Instructions encode the number of most-significant mantissa bits to be computed in floating point operations. They are typically used to achieve execution time and energy reductions. First experiments using the tool show promising results.