Task Assignment for Network Processor Pipelines Using GA
In several commercial network processors programming environments, programmer must manually assign many processing tasks to the processor pipelines which consist of many processing engines. Due to the large exploration space, this manual procedure is usually very tedious and inefficient and the optimal or even near-optimal assignment scheme may be difficult to obtain. This paper proposes an automated task-to-PE assignment algorithm based on genetic algorithm. Experimental results show that this method can quickly obtain near-optimal solutions from the large solution space and the algorithm execution time is decoupled with pipeline stages. These two features make this method very suitable to be used in a NP application development environment and provide a more efficient development experience for developers.
Unable to display preview. Download preview PDF.
- 1.Xiaoping, W., Liming, C.: Genetic Algorithm: Theory, Application and Software Implementation, Xi’An. Press of Jiaotong UniversityGoogle Scholar
- 2.Franklin, M.A., Datar, S.: Pipeline Task Scheduling on Network Processors. In: Workshop on Network Processors & Applications - NP3, Madrid, Spain (Febuary 2004)Google Scholar
- 4.Goldberg, D.E.: Genetic Algorithms in Search, Optimization, and Machine Learning (Hardcover). Addison-Wesley Professional, Reading (January 1989)Google Scholar
- 5.Weng, N., Wolf, T.: Pipelining vs. Multiprocessors – Choosing the Right Network Processor System Topology. In: Proceedings of ANCHOR 2004 (June 2004)Google Scholar
- 6.Dai, J., Huang, B., et al.: Automatically Partitioning Packet Processing Applications for Pipelined Architectures. In: Proceedings of ACM SIGPLAN PLDI 2005 (June 2005)Google Scholar
- 7.Shoumeng, Y., Xingshe, Z., Lingmin, W.: GA-Based Automated Task Assignment on Network Processors. In: Proceedings of ICPADS 2005 (July 2005)Google Scholar
- 8.Fan, Z., Xingshe, Z., Shoumeng, Y.: Design and Implementation of Network Processor Programming Model Based on Software Component. In: To appear in Computer Engineering and Applications (in Chinese) Google Scholar