一种基于双模结构的预测器

A Predictor Basing on Two-Pattern Structure

分支预测是限制微处理器性能提高的一个重要因素,因此也一直是微处理器设计研究的重点.文中提出了一种对动态两级自适应分支预测进行改进的新方法,即基于双模结构的双模预测器.它能同时根据指令转移间相关和转移内相关进行转移预测,从而获得更高的转移预测精度.

Branch prediction is an important aspect to improve the performance of microprocessor, and it is one of the most important field in the research of microprocessor. This paper proposes a new method to improve the Two-Level Adaptive Training branch predictor, which is called the two-pattern predictor basing on a two-pattern structure. It makes predictions based on both inter-branch correlation and intra-branch correlation simultaneously and then can achieve higher branch prediction accuracy.