2026-08-08T04:31:27Z https://drops.dagstuhl.de/oai

oai:drops-oai.dagstuhl.de:20471 2024-07-26T08:47:22Z ddc:004 open_access

Invited Paper: Assessing Unchecked Factors for Certification: An Experimental Approach for GPU Cache Parameters Cazanove, Cédric Lesage, Benjamin Boniol, Frédéric Ermont, Jérôme GPU benchmarks simulation certification The certification objectives for airborne electronic hardware defined in AMC20-152A [EASA, 2021] and in AMC20-193 [EASA, 2020] capture some of the activities required for an applicant to embed a hardware platform in a safety-critical avionic system. For COTS (Commercially available Off-The-Shelf) platforms in particular, these objectives require applicants to identify functions, configuration settings, and resources present on the platform, and assess their use by the system. AMC20-152A however recognizes that documentation regarding the behavior of a COTS may be incomplete. There is thus a strong push for applicants to the certification of a COTS to demonstrate their mastery of the platform, to highlight relevant factors (functions, settings, resources, etc.), and their use in their system. We outline in the following a standard approach to the exploration of unchecked factors of a platform, considering existing approaches in the literature, to build such a mastery. Our approach incrementally incorporates and validates knowledge of various factors by including them in micro-simulations compared to experimental ground truth. Schloss Dagstuhl – Leibniz-Zentrum für Informatik Cédric Cazanove and Benjamin Lesage and Frédéric Boniol and Jérôme Ermont 2024 Is Part Of OASIcs, Volume 121, 22nd International Workshop on Worst-Case Execution Time Analysis (WCET 2024) InProceedings Text doc-type:ResearchArticle publishedVersion application/pdf doi:10.4230/OASIcs.WCET.2024.3 urn:nbn:de:0030-drops-204719 https://drops.dagstuhl.de/entities/document/10.4230/OASIcs.WCET.2024.3 eng https://creativecommons.org/licenses/by/4.0/legalcode