Author retrospective improving data cache performance by pre-executing instructions under a cache miss
Author retrospective improving data cache performance by pre-executing instructions under a cache miss
复制标题
作者回顾通过在缓存未命中时预执行指令来提高数据缓存性能
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
T. Mudge
中科院分区：
文献类型：
--
作者：
T. Mudge
本文介绍并评估了一种称为 runahead 的技术，该技术可在缓存未命中时将指令和数据预取到 L1 缓存中。论文中报道的实验结果表明，简单有序管道的 CPI 可以降低 30%。与许多其他预取策略相比，提前运行的主要优点有两个：1) 它预取指令和日期； 2) 它能够创建非顺序数据预取。原始论文：http://dx.doi.org/10.1145/263580.263597。
The paper introduces and evaluates a technique, referred to as runahead, that prefetches instructions and data into the L1 caches on cache misses. The results of the experiments reported in the paper show that the CPI of a simple in-order pipeline could be reduced by 30%. The principle advantage of runahead over many other prefetching strategies, is twofold: 1) it prefetches both instructions and date; and 2) it is able to create non-sequential data prefetches. Original Paper: http://dx.doi.org/10.1145/263580.263597.