Dataset - Automated Design Space Exploration of Approximation-Enabled Accelerators: A Systematic Literature Mapping

Data collected for systematic literature mapping on design space exploration strategies for approximate computing accelerators.

Abstract— This paper provides a systematic mapping ofthe literature regarding automated Design Space Exploration(DSE) for error-tolerant systems, where Approximate Com-puting (AxC) is employed to provide gains in energy efficiencyor performance. Starting from the Scopus aggregator andIEEE Xplore library, 1302 texts were analyzed, from which 43peer-reviewed research works were selected to map the state-of-the-art in DSE techniques for designing approximate hard-ware accelerators. The selected papers were used to catalogthe most commonly used search strategies, the methods for in-tegrating approximate units and techniques in the design spaceof a given system, and challenges and opportunities for the us-age of automated DSE in AxC-enabled workflows. The resultshighlight the usage of genetic and evolutionary algorithms, andidentify a trend for tree search-based exploration, with a focuson selection of arithmetic units and functional approximation.This study also shows the challenge of integrating a DSE stagein existing workflows, particularly due to the difficulty of an-alyzing error and electrical characteristics with feasible run-times.