摘要
值预测和指令重用是通过开发程序执行结果的冗余来解决数据相关的两种不同的新技术。本文首先从这两种新技术的原理出发,深入剖析了它们的技术特性,然后研究了它们与微体系结构其它特征间的相互影响,最后评估了这些技术对微处理器性能的影响。
Value Prediction (VP) and Instruction Reuse (IR) are two novel techniques that have recently been proposed for exploiting the redundancy in programs to collapse data dependence. In this paper, we first differentiate the two techniques from their principles. Furthermore, we qualitatively discuss their interaction with other microarchitectural features. Finally, we evaluate the impact of these techniques on the performance of microprocessors.
出处
《计算机工程与科学》
CSCD
2005年第11期98-101,共4页
Computer Engineering & Science
基金
国家自然科学基金资助项目(60273069)