2-Bit Branch Predictor Modeling Using Markov Model

Power consumption is a very important issue when it comes to embedded devices, therefore every processing cycle should be optimally utilized and considered. In speculated execution, highly mispredicted branches are considered a critical threat for both time and power saving. In this paper, we show that, for a specific branch, misprediction rate of a 2-bit branch predictor can be precisely calculated using Markov model. Further, this can be done offline for more power saving. Thus, a decision of replacing the branch with conditional (predicated) instructions instead of counting on the predictor can be made.