In this project, the main goal is development of a simulator for modeling a global branch predictor. For a global branch predictor, separate history records are not kept for different conditional jumps. In fact, a shared history record is maintained for all of the conditional jumps.

With this feature, any correlation between different conditional jumps helps in making more accurate predictions. However, the shared history record may be filled with useless information and the needed information for a branch may be lost due to limited space. The global branch predictor can use a two$-$level adaptive mechanism. This scheme is beneficial only for large table sizes. The size of the pattern history table has exponential relationship with the size of the history buffer. The pattern history table should be large enough to be shared among all conditional jumps. There are two types for a two$-$level adaptive predictor with globally shared history buffer and pattern history table: $($a$)$ gshare, when it performs XOR operation on the global history and branch program