Enhancing Branch Predictors using Genetic Algorithm
Md Sarwar M Haque, Md. Rafiul Hassan, Muhammad Sulaiman, Salami Onoruoiza, Joarder Kamruzzaman, Md Arifuzzaman · 2019
Dynamic branch prediction is a hardware technique used to speculate the direction of control branches. Inaccurate prediction will make all speculative works useless while accurate prediction will significantly improve microprocessors performance. In this work, we have shown that Genetic Algorithm (GA) can be used to select (near) optimal parameters for branch predictors in most cases. The GA-enhanced predictors take time to find suitable parameters, but once the values of these parameters are determined, the GA-enhanced predictors take the same time to execute as the basic predictors with increased accuracy.