遴选算法全局条件解耦与物理优化实现

Global Constraints Decoupling and Physical Optimal Implementation of Expert-chosen Algorithm

针对遴选算法处理信息量大、条件复杂、运行时间长等特性,研究算法准并行化的方法,运用全局约束量主动推送和全局记录加锁策略对全局条件进行解耦,在此基础上建立一种基于客户端/服务器模式的多线程算法结构,并对2种解决并行同步问题的加锁策略进行比较分析。实验结果表明,优化实现后算法的运行速度有明显提升。

Expert-chosen algorithm need process a large amount of data under complex constraints, it is a time consuming work. This paper examines the quasi-parallel approach of the algorithm. The strategies of global constraints status automatic delivering and global record locking are used to decouple the global constraints. Based on this, this paper proposes a client/server multi-threading architecture and conducts a comparative analysis for two locking methods which handles the synchronization. Experimental results show that the running speed of the optimizing algorithm is significantly improved.