Hybrid Predictor Unit Modeling

This work is devoted to the development of a hybrid transition predictor, which includes several transition prediction algorithms. The work includes a review of existing prediction algorithms, such as static and dynamic prediction, which in turn is divided into many different algorithms such as local or global prediction. Analysis of existing technical solutions in the field of development of hybrid transition predictors showed that this industry is by far the most promising. In addition, a study was conducted of vulnerabilities that use the introduction of branching predictors as a weak point of the processor. A prototype hybrid transition predictor was developed, which includes three dynamic prediction algorithms, and then its modification was carried out, which made it possible to achieve prediction accuracy of more than 85%. The volume of work is 46545 characters, contains 11 figures, 21 sources.