Theoretical Analysis of Branch Prediction in the Sliding Window Algorithm
Résumé
We revisit the classical sliding window pattern matching algorithm from a computer architecture point of view and conduct a theoretical study to answer the question: what is the impact of adding a simple branch prediction mechanism to the model of computation? Assuming that the pattern is fixed and the text is random, we give precise estimations of the number of mispredictions produced by two variants of this algorithm using local predictors. This case study introduces key tools based on automata and Markov chain theories, which are promising for the analysis of more efficient algorithms and more advanced branch prediction schemes.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |