Loading...

A New Algorithm for Partitioning Problem in Hardware/Software Codesign

Tahaee, Abdoreza | 2009

623 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 41584 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Jahangir, Amir Hossein
  7. Abstract:
  8. We use an abstract model to perform hardware/software partitioning for embedded systems. The complexity of partitioning problem in the model is studied and we use a scheme to overcome its NP-Hardness. Partitioning problem can be considered as the main phase of high level synthesis. As it suffers from complexity, heuristic search methods are generally used to overcome its complexity. In this thesis we present a heuristic method (named Tendency algorithm) to increase the performance of heuristic searches, and use it in parallel with other tuning methods. We defind suitable initial point(s) for heuristic searches instead of tuning heuristic parameters. The thesis justifies the logic behind tendency algorithm by measuring the “entropy” of solution (in regard to optimal solution), and by numerous empirical tests. It shows that the produced initial points are closer to global optimum than random points, or even points that are produced by pre-running Genetic search.In progress, the thesis takes a more theoretical approach to focus on algorithmic properties of hardware/software partitioning. This makes the core of PhD thesis. We introduce several theorems in this scope of knowledge. A method is proposed to find or approach the global optimum of an NP-Hard model of partitioning problem in some of its instances. We approach with lower bound distance to global optimum in the remaining of problem instances. The strategy is based on smartly changing the parameters of polynomial model of partitioning problem while forcing it to produce (or to approach) the exact answer of NP-Hard model. The capability of method to solve other hard problems is also studied.
  9. Keywords:
  10. Genetic Algorithm ; Hardware-Software Co-Design ; Maximum Flow ; Hardware-Software Partitioning ; NP-Hard Problems ; Minimum Cut

 Digital Object List

 Bookmark

No TOC