Implemented Static Branch Prediction Schemes for the Parallelism Processors
Sweety Nain, Prachi Chaudhary · 2019
Branch prediction technique is a vital part of modern processors. An instruction level parallelism (ILP) processor uses prediction schemes to execute the instructions along with the predicted path. When the pipeline gets deeper, the behavior of branch instructions critically affects the performance of parallelism processors and the branch misprediction rates are major obstacles to achieving higher performance of pipeline processors. This paper presents are an insight in to analysis and comparison of static branch prediction techniques such as; always taken prediction scheme and always not-taken prediction scheme. In this paper, these techniques are compared on the basis of the branch address hit accuracy rate and branch misprediction rate by using SPEC-alpha benchmarks. While comparing, the highest address accuracy rate obtained is 75.75% in always taken prediction scheme as compared to 66.40% of always not-taken prediction schemes and least branch misprediction rate obtained is 24.24% in always taken prediction schemes as compared to 33.07% of always not-taken prediction schemes.