{"title":"High performance branch prediction","authors":"B. M. Guy, R. Haggard","doi":"10.1109/SSST.1996.493550","DOIUrl":null,"url":null,"abstract":"Today's advanced architectures increasingly rely on accurate instruction fetch and branch prediction to maintain optimum pipeline performance. In this paper, dynamic branch prediction techniques are investigated and methods to increase branch prediction accuracy are explored. A result of this analysis has been the development of several dynamic branch prediction architectures that combine promising aspects of accurate branch prediction methods currently implemented. The objective of this research was to provide an advanced high performance branch prediction architecture that can easily be incorporated into existing architectures without instruction set modification. Development of a hybrid branch prediction architecture that uses both global and local branch histories is presented and shown to have significant performance advantages. Several branch architectures are presented for performance.","PeriodicalId":135973,"journal":{"name":"Proceedings of 28th Southeastern Symposium on System Theory","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1996-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of 28th Southeastern Symposium on System Theory","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/SSST.1996.493550","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
Today's advanced architectures increasingly rely on accurate instruction fetch and branch prediction to maintain optimum pipeline performance. In this paper, dynamic branch prediction techniques are investigated and methods to increase branch prediction accuracy are explored. A result of this analysis has been the development of several dynamic branch prediction architectures that combine promising aspects of accurate branch prediction methods currently implemented. The objective of this research was to provide an advanced high performance branch prediction architecture that can easily be incorporated into existing architectures without instruction set modification. Development of a hybrid branch prediction architecture that uses both global and local branch histories is presented and shown to have significant performance advantages. Several branch architectures are presented for performance.