Alternative Implementations for Hybrid Branch Predictors

Po-Yung Chang, Eric Hao, Yale N. Patt

Abstract

Highly accurate branch prediction is an important requirement for achieving high performance on deeply pipelined, superscalar processors. To improve on the prediction accuracy of current single-scheme branch predictors, hybrid (multiple-scheme) branch predictors have been proposed~\cite{mcf93,cha:hao94}. These predictors combine multiple single-scheme predictors into a single predictor. They use a selection mechanism to decide for each branch, which single-scheme predictor to use. The performance for a given implementation of a hybrid predictor is dependent on the single-scheme predictors used and the selection mechanism. To find the most effective implementations, this paper examines several hybrid predictor implementations and compares their prediction accuracies. In addition, it introduces a new selection mechanism, the 2-level selector, which improves the performance of the hybrid branch predictor.

Keywords

branch prediction, speculative execution, superscalar.

Talk Overheads (329764 bytes)