Performance of parallel consistency algorithms
Consistency algorithms are used to reduce the overall computation time for consistent labeling problem which is commonly used in computer vision. We have given parallel algorithms to enforce arc consistency, which has been shown to be inherently sequential[3,6]. Since all three parallel algorithms have the same time complexity on a PRAM, it is necessary to implement them on a parallel processor to determine which one performs best. The results of implementing them on a BBN Butterfly are presented here.
KeywordsRandom Graph Sequential Algorithm Parallel Processor Binary Constraint Cyclic Graph
Unable to display preview. Download preview PDF.
- John Gaschnig. Performance Measurements and Analysis of Certain Search Algorithms. PhD thesis, Carnegie-Mellon University, Department of Computer Science, May 1979.Google Scholar
- Simon Kasif. On the parallel complexity of some constraint satisfaction problems. In Proceedings of AAAI-86, pages 349–353. AAAI, August 11–15 1986.Google Scholar
- Ashok Samal. Parallel Split-Level Relaxation. PhD thesis, Department Computer Science, University of Utah, August 1988.Google Scholar