摘要
分支预测技术一直以来是计算机体系结构、微处理器设计的研究重点。目前分支预测的研究集中在动态分支预测技术,采用学科交叉的观点,提出新的预测算法。对自适应动态分支预测进行改进,引入了模糊加权的机制,对分支历史的每一位加不同的权值,并利用调整因子动态改变权值,由模糊推理得出预测结果。SimpleScalar的模拟结果表明,这种模糊加权的动态自适应算法比经典的gshare预测算法预测失效率低2%。
Branch prediction technology is one of the most important field in the research of computer architecture and microprocessor. The research on branch prediction focuses on dynamic branch prediction. Many new prediction algorithms have been raised based on the merge of different subject, The adaptive dynamic branch prediction mechanism has been improved and fuzzy weight mechanism has been introduced. Every bit of the BHR has a different weight and dynamically changed by a adjust gene. The prediction result was given through fuzzy consequence. The result of SimpleScalar simulation shows that the miss prediction rate of dynamic branch prediction algorithm based on fuzzy weight is 2% lower than the classical gshare prediction mechanism.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2006年第12期3390-3392,共3页
Journal of System Simulation
基金
自然科学基金项目(60376018)
关键词
分支预测
模糊加权
调整因子
模拟器
Branch Prediction
Fuzzy Weight
Adjust Gene
Simulation
