GAROP: Genetic Algorithm framework for Running On Parallel environments

概要

論文の詳細を見る
In this research, a Genetic Algorithms framework for Running On Parallel environments, which is named GAROP, is proposed. The GAROP provides the library for a parallel processing, so that users should only describe codes for genetic algorithms (GA) programs, utilizing the library implemented for the part requiring a parallel processing. In the GAROP framework, GA research provides only program codes which are concerned with GA algorithm and GAROP library supports other codes which are concerned with parallel processing. The advantage of using GAROP is to increase the user's productivity by making it possible to develop the program, which can execute a parallel processing. In this paper, the broad description of the GAROP is provided, and the development of the GAROP, corresponding multi-core CPU and GPU environments, is described. The libraries are implemented with GA which finds quasi-optimum solutions using meta heuristics, and its productivity and its parallelism are evaluated. As a result, only adding four descriptions to the program, the acceleration of the processing speed is confirmed in both of the environments; 5.26 times speed-up on multi-core CPU, and 3.0 times speed-up on GPU.
2012-07-09