A novel meta predictor design for hybrid branch prediction

Abstract

Recent systems have been paved the way for being high-performance due to the super-pipelining, dynamic scheduling and superscalar processor technologies. The performance of the system is greatly affected by the accuracy of the branch prediction because the overhead of each misprediction has grown due to greater number of instructions per cycle and the deepened pipeline. Hybrid branch prediction is usually used to increase the prediction accuracy on such high-performance systems. Normally hybrid branch prediction uses several branch predictors. A meta-predictor selects which branch predictor should be used corresponding to the program context of the branch instruction instance for the branch prediction. In this paper, we discuss about the saturating counter within meta predictor. The design of the saturating counter which selects a predictor that has high-prediction ratio has brought out the high accuracy of the prediction for the branch predictor.