분기 연관성을 이용한 분기 예측기 설계

This paper proposed a new efficient branch predictor with branch correlation by BRDT to increase performance of microprocessor with speculative superscalar structure and deep pipelining. Proposed branch predictor model consists of Bimodal predictor with 1K fixed PHT size as Local predictor and a-share predictor s Global predictor. Checking simulation result, proposed branch predictor model with 4K global PHT size achieved 1.7% higher accuracy at prediction rate on average. Also proposed model with 2K of PHT showed much higher prediction rate than original g-share predictor with 4K of PHT. So proposed model achieves high performance per area.