基于投机执行的两级退休机制

发布时间：2018-01-08 14:14

本文关键词：基于投机执行的两级退休机制　出处：《计算机应用研究》2015年04期 　论文类型：期刊论文

【摘要】：针对超标量处理器中指令长时间占用重排序缓存引起指令退休缓慢的问题,提出了一种基于投机执行的两级退休机制。该方案根据指令有无异常和预测错误风险将指令分为有风险指令和无风险指令,对重排序缓存进行轻量化改进,只有存在异常和预测风险的指令才允许进重排序缓存,在确认风险消除后将指令快速退休。重命名寄存器从重排序缓存分离,负责寄存器重命名和结果乱序回写。实验结果表明,在硬件资源相同的情况下,基于该方案的处理器比传统的按序退休处理器的性能平均提高28.8%以上。
[Abstract]:In superscalar processors, the problem of slow instruction retirement caused by the long time occupation of reorder cache in superscalar processors is discussed. This paper proposes a two-level retirement mechanism based on speculative execution. The scheme classifies instruction into risk-free instruction and risk-free instruction according to the risk of instruction anomaly and prediction error and improves the reorder cache lightweight. Only instructions with exceptions and prediction risks are allowed to reorder the cache, and the instructions are retired quickly after confirming the risk. The rename register is separated from the resort cache. The experimental results show that the performance of the processor based on this scheme is more than 28.8% higher than that of the traditional sequential retirement processor under the same hardware resources.
【作者单位】：浙江大学电气工程学院;浙江大学信息与电子工程学系;
【基金】：国家科技重大专项核高基重大专项资助项目(2009ZX01030-001-002)
【分类号】：TP332
【正文快照】： 0引言随着科学技术的快速发展,嵌入式系统的应用越来越广泛,对嵌入式处理器的性能要求越来越高。在超标量乱序执行的嵌入式处理器中,重排序缓存(reorder buffer,ROB)负责维护精确异常、硬件推测和寄存器重命名等功能。为了保证处理器的效率,重排序缓存的数目必须大于流水线中

【参考文献】