摘要
随着高性能微处理器的流水线越来越深,分支预测的精度不够高已经成为制约微处理器性能进一步提高的瓶颈。由于一种分支预测方式总会有一些缺陷,所以复合分支预测方式就被提出来。也就是利用几种分支预测方式同时进行预测,再基于某种选择算法选择采用哪一种预测方式的结果。基于这种复合分支预测方式,提出了几种新的选择算法实现。
With the deeper and deeper of the pipeline , the inadequate accuracy of branch prediction has become the bottleneck for the further elevation of the microprocessor performance . As a mechanism of branch prediction always has limitation . A hybrid branch prediction was presented. The predict mechanism uses several mechanisms of branch prediction ,then select a result of them based on a special selection algorithm. Presents several implementations of the new selection arithmetic.
出处
《计算机应用研究》
CSCD
北大核心
2005年第4期27-29,共3页
Application Research of Computers
基金
国家自然科学基金资助项目(60376018)
关键词
复合分支预测
选择算法
权
汉明距离
Hybrid Branch Prediction
Selection Algorithm
Weight
Hamming Distance