Branch prediction is crucial to maintaining high performance in modern Superscalar processor. Today's Superscalar processors achieve high performance by executing multiple independent instructions in parallel. One of the most impedement to the performance of wide-issue superscalar processor is the presence of conditional branches. Conditional branches can occur as frequently as one in every 5 or 6 instructions, leading to heavy misprediction penalties in superscalar architectures. Ideal speed-up in superscalar processor is seldom achieved due to stalls and breaks in the execution stream. These interrupts are caused by data and control hazards which deteroits the superscalar processor performance. Branch target buffer (BTB) can reduces the performance penalty of branches in superscalar processor by predicting the path of the branch and caching information used by the branch. No stalls will be encountered if the branch entry is found in the BTB and prediction is correct. Otherwise, the penalty will be of atleast '2' cycles. This paper proposes an algorithm for superscalar processor based on changing the BTB structure to eliminate the misprediction penalty. It also highlights a problem in the previous BTB algorithm (nested branches problem) and proposes a solution to it.
This paper proposes modified architecture of 21264, Out-Of-Order, six-way issue microprocessor. The proposed modified architecture implements Tomasulo's algorithm using tournament branch prediction scheme to improve the performance of processor. Tomasulo's Algorithm controls the operation of the Common Data Bus (CDB) by means of tag mechanism. A tag is a 4-bit number used to identify separately each of eleven sources which can feed the CDB. The proposed modified architecture will evaluate branch outcome by taking both local and global history. The choice of global-versus-local branch prediction is made dynamically on a path-based predictor that decides which predictor to use, based on the past correctness of choice. Index Terms-common data bus (CDB), tomasulo's algorithm, tournament branch predictor.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.