通过SimpleScalar的分支预测模型设计与研究

A Design and Research of Branch Prediction Model Based on SimpleScalar

现如今,计算机体系结构是计算机研究领域的热点。使用传统的硬件描述语言(Hardware Description Language,HDL)构建体系结构时,需要大量的时间和硬件资源,而处理器模拟器能够快速地构建实验原型,辅助计算机体系结构的探索。本文通过在Simple Scalar模拟器的分支预测模块中添加了混合(Hybird)分支预测方法以及感知器(Perceptron)预测方法,利用SPEC95INT中的测试集进行了实验,并对不同的分支预测模型实验结果进行了分析,为实际应用中的微处理器分支预测器的设计提供了参考。

Nowadays,computer architecture is a hotspot in the field of computer research.It takes a lot of time and hardware resources to build the architecture with the traditional hardware description language(HDL),but the processor simulator can quickly build experimental prototypes to assist the exploration of computer architecture.In this paper,a hybrid branch prediction method and a Perceptron prediction method are added to the branch prediction module of the SimpleScalar simulator,and the experiment is carried out by using the test set in SPEC95INT.The experimental results of different branch prediction models are analyzed,which provides a reference for the design of microprocessor branch predictor in practical application.