A Hybrid Parallel Genetic Algorithm with Dynamic Migration Strategy Based on Sunway Many-Core Processor
Yao Liu, Ruixiang Zhao, Kai Zheng, Su Wang, Yan Liu, Huanxue Shen, Qianhao Zhou · 2017
To speed up the execution of Genetic Algorithm when compute-intensive tasks are encountered, the Hybrid Parallel Genetic Algorithm with Dynamic Migration strategy (HPGA-DM) is designed and implemented on Sunway TaihuLight supercomputer which is ranked the first in the latest TOP500 list. Different from the traditional Genetic Algorithm implemented on single-core or multi-core cluster with single-level parallel architecture, the algorithm using two-level parallel architecture which combines with two different programming models: MPI and Athread, and is implemented on the Sunway Many-Core Processor. At the same time, a dynamic migration strategy is presented to improve the convergence speed of the algorithm. The experimental results show that HPGA-DM is 20.2 times faster than the single-level parallel algorithm, and obtains the 32.24 times slave-core speedup and 50.37% execution efficiency when the number of CPEs is 16*64.