GPU Accelerating Statistical Model Checking for Extended Timed Automata

Oliver S. Bak, Mathias W.B. Christiansen, Oliver V. Eriksen, Sergio Feo-Arenis, Peter G. Jensen, Marcus D. Jensen, Simas Juozapaitis, Kim G. Larsen, Marius Mikučionis*, Marco Muñiz, Andreas Windfeld

*Kontaktforfatter

Publikation: Bidrag til bog/antologi/rapport/konference proceedingBidrag til bog/antologiForskningpeer review

Abstract

The core component of Statistical Model Checking (SMC) is the repeated sampling of a given system as to estimate statistical measures. To obtain probabilistic estimates with high confidence a significant number of simulations is required, in particular in the presence of rare events. In this paper we explore the use of Graphical Processing Unit (GPU) for accelerating SMC for Networks of Stochastic Extended Timed Automata (SXTA). We discuss the many challenges and solutions required to achieve significant speedups on a GPU architecture. In collaboration with NVIDIA we develop a prototype tool for parallel SMC using both GPU and multi-core CPU. Experimental results demonstrate trade-offs in the computation time when utilizing either CPU or GPU. In one case we observed the GPU using 20% of the power of the CPU equivalent while delivering a 2.73 time speedup.

OriginalsprogEngelsk
TitelLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
RedaktørerNils Jansen, Sebastian Junges, Benjamin Lucien Kaminski, Christoph Matheja, Thomas Noll, Tim Quatmann, Mariëlle Stoelinga, Matthias Volk
Antal sider26
Vol/bind15261
ForlagSpringer Science+Business Media
Publikationsdato2025
Sider267-292
ISBN (Trykt)978-3-031-75774-7
ISBN (Elektronisk)978-3-031-75775-4
DOI
StatusUdgivet - 2025
NavnLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Vol/bind15261 LNCS
ISSN0302-9743

Bibliografisk note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Fingeraftryk

Dyk ned i forskningsemnerne om 'GPU Accelerating Statistical Model Checking for Extended Timed Automata'. Sammen danner de et unikt fingeraftryk.

Citationsformater