Dynamic Verification Framework of Approximate Computing Circuits using Quality-Aware Coverage-Based Grey-Box Fuzzing

摘要

Approximate computing (AC) has recently emerged as apromising approach to the energy-efficient design of digital systems. Forrealizing the practical AC design, we need to verify whether the designedcircuit can operate correctly under various operating conditions. Namely,the verification needs to efficiently find fatal logic errors or timing errorsthat violate the constraint of computational quality. This work focuses onthe verification where the computational results can be observed, the computationalquality can be calculated from computational results, and theconstraint of computational quality is given and defined as the constraintwhich is set to the computational quality of designed AC circuit with givenworkloads. Then, this paper proposes a novel dynamic verification frameworkof the AC circuit. The key idea of the proposed framework is toincorporate a quality assessment capability into the Coverage-based GreyboxFuzzing (CGF). CGF is one of the most promising techniques in theresearch field of software security testing. By repeating (1) mutation of testpatterns, (2) execution of the program under test (PUT), and (3) aggregationof coverage information and feedback to the next test pattern generation,CGF can explore the verification space quickly and automatically. On theother hand, CGF originally cannot consider the computational quality byitself. For overcoming this quality unawareness inCGF, the proposed frameworkadditionally embeds the Design Under Verification (DUV) componentinto the calculation part of computational quality. Thanks to the DUV integration,the proposed framework realizes the quality-aware feedback loopin CGF and thus quickly enhances the verification coverage for test patternsthat violate the quality constraint. In this work, we quantitatively comparedthe verification coverage of the approximate arithmetic circuits between theproposed framework and the random test. In a case study of an approximatemultiply-accumulate (MAC) unit, we experimentally confirmed thatthe proposed framework achieved 3.85 to 10.36 times higher coverage thanthe random test.