On Parallel Implementations of Deterministic Finite Automata

  • Authors:
  • Jan Holub;Stanislav Štekr

  • Affiliations:
  • Department of Computer Science and Engineering, Faculty of Electrical Engineering, Czech Technical University in Prague, Prague 2, Czech Republic 121 35;Department of Computer Science and Engineering, Faculty of Electrical Engineering, Czech Technical University in Prague, Prague 2, Czech Republic 121 35

  • Venue:
  • CIAA '09 Proceedings of the 14th International Conference on Implementation and Application of Automata
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present implementations of parallel DFA run methods and find whether and under what conditions is worthy to use the parallel methods of simulation of run of finite automata. First, we introduce the parallel DFA run methods for general DFA, which are universal, but due to the dependency of simulation time on the number of states |Q | of automaton being run, they are suitable only for run of automata with the smaller number of states. Then we show that if we apply some restrictions to properties of automata being run, we can reach the linear speedup compared to the sequential simulation method. We designed methods benefiting from k -locality that allows optimum parallel run of exact and approximate pattern matching automata. Finally, we show the results of experiments conducted on two types of parallel computers (Cluster of workstations and Symmetric shared-memory multiprocessors).