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.
Original language | English |
---|---|
Title of host publication | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
Editors | Nils Jansen, Sebastian Junges, Benjamin Lucien Kaminski, Christoph Matheja, Thomas Noll, Tim Quatmann, Mariëlle Stoelinga, Matthias Volk |
Number of pages | 26 |
Volume | 15261 |
Publisher | Springer Science+Business Media |
Publication date | 2025 |
Pages | 267-292 |
ISBN (Print) | 978-3-031-75774-7 |
ISBN (Electronic) | 978-3-031-75775-4 |
DOIs | |
Publication status | Published - 2025 |
Series | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
Volume | 15261 LNCS |
ISSN | 0302-9743 |
Bibliographical note
Publisher Copyright:© The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.