Mehdi Bouaziz
Scalable Numerical Abstact Domains
2nd Workshop on Analysis and Verification of Dependable Cyber Physical Software (AVDCPS 2013), National University of Defense Technology, Changsha, Chine. 23 novembre 2013
- PDF, 64 pages, 472 Ko
- Résumé : Except from intervals, numerical abstract domains do not scale up. When it comes to analyzing large programs, abstract interpretation-based tools must ensure near-linear time operations. This is generally achieved by splitting the set of variables into independent smaller sets, sometimes sharing some non-relational information. We present a method to increase precision whilst retaining a linear complexity ensuring scalability. Suited for all subpolyhedral domains, the method consists in organizing the subsets of variables, sharing fully expressive relations, into a tree-shaped graph allowing an efficient completion algorithm to be performed.
- @misc{Bouaziz-AVDCPS13-talk,
author = {Bouaziz, Mehdi},
title = {Scalable Numerical Abstact Domains},
year = {2013},
month = {November},
booktitle = {2nd Workshop on Analysis and Verification of Dependable Cyber Physical Software (AVDCPS 2013)},
institution = {National University of Defense Technology},
url = {http://www.bouaziz.me/p/Bouaziz-AVDCPS13-slides.pdf}
}
Dernière mise à jour : 24 janvier 2014