首页> 外国专利> Quick index tree for accelerated branching forecast

Quick index tree for accelerated branching forecast

机译:快速索引树,用于加速分支预测

摘要

Advanced branching prediction system using a fast index tree (FIT) with the system having:a branching buffer;a FIT structure, comprising a table or tables, which index the positions of entries in the branching target buffer, with one or more tables having a majority of entries;of which each represents a node in a binary tree and specifies a first next predicted branching in the branching target buffer along a missed branching path and a second predicted branching in the branching target buffer along a selected branching path,where each node of the binary tree represents a branching in a program code;anda processing circuit which is configured to perform a process comprising:Determine that the branching buffer shall be searched under FIT control;Save a current search address to search the branch target buffer;Search the branch target buffer at the saved current search address for a branch forecast;Determine a next FIT search address based on reading branching paths taken and not taken from the FIT structure for a next search stage of predicted branching;re-indexing the branching buffer search based on the next FIT search address;Determine whether the branch forecast has been found by searching the saved current search address;Based on the fact that the branch forecast was found by searching the saved current search address, continue searching the branch target buffer under FIT control, setting the current search address based on the next FIT search address;andbased on the fact that the branching forecast was not found by searching the saved current search address, re-indexing the browsing of the branching target buffer with the saved current search address and performing the browsing of the branching target buffer without FIT control.
机译:使用快速索引树(适合)具有系统具有:分支缓冲区的高级分支预测系统;一个拟合结构,包括表或表,其索引分支目标缓冲区中的条目位置,其中一个或多个表其中大多数条目;其中每个条目表示二进制树中的节点,并且沿着所选择的分支路径,沿着所选分支路径在分支目标缓冲器中指定分支目标缓冲器中的第一个接下来预测分支,其中每个节点二进制树代表程序代码中的分支; Ada处理电路被配置为执行包括:在拟合控件下确定分支缓冲区应搜索分支缓冲区;保存当前搜索地址以搜索分支目标缓冲区;搜索分支预测的已保存当前搜索地址的分支目标缓冲区;基于读取分支路径确定下一个适合搜索地址,而不是拍摄ROM适用于预测分支的下一个搜索阶段的拟合结构;基于下一个适合搜索地址重新索引分支缓冲区搜索;通过搜索所保存的当前搜索地址来确定是否已经找到了分支预测;基于该事实通过搜索保存的当前搜索地址来搜索分支预测,继续在拟合控件下搜索分支目标缓冲区,根据下一个适合搜索地址设置当前搜索地址;并基于搜索保存的分支预测当前搜索地址,重新索引分支目标缓冲区的浏览,并使用已保存的当前搜索地址并执行浏览分支目标缓冲器而不适合控制。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号