A Low-power Design for gshare Branch Predictor
Wu Meng, Haibin Shen
Abstract
Wu Meng, Haibin Shen
Abstract
Power and performance are both the important design issues in high-end embedded computing system. Based on the microarchitecture of dynamic branch predictor used in deep-pipelined processor, this paper proposes a method to reduce the power of gshare branch predictor using banking technique and explores the relationship among predictor's size, number of banks and predictor's area/power.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Power and performance are both the important design issues in high-end embedded computing system. Based on the microarchitecture of dynamic branch predictor used in deep-pipelined processor, this paper proposes a method to reduce the power of gshare branch predictor using banking technique and explores the relationship among predictor's size, number of banks and predictor's area/power.
Key concepts: Branch predictor, Computer science, Microarchitecture, Power (physics), Predictive power, Parallel computing, Embedded system, Computer architecture