中文摘要：

随着工艺尺寸缩小和处理器频率的提高，大容量的片上L2cache成为处理器漏流功耗的主要来源。提出的保守多状态（C-SP＆SD）和推断多状态S-SP＆SD）两种L2 cache漏流功耗控制策略能够将状态保留（State-Preserving）与状态破坏（State-Destroying）两种低功耗模式相结合。如果一个数据在多级cache存储层次中存在多个副本，那么只保留一个副本处于活跃状态，其他副本均被转换到低功耗模式，并且在不显著影响处理器性能的前提下尽可能转换到更低功耗的状态破坏模式。与传统的L2 cache漏流控制策略相比，C-SP＆SD策略以较小的处理器性能损失换取较大的L2 cache漏流功耗节省，而S-SP＆SD策略则实现了最优的L2 cache漏流功耗节省和处理器能量效率。

英文摘要：

As featme size shrinks and the frequency increases,on-chip L2 cache with large capability is the main source of leakage power in micro-processors. Supporting not only the State-Preserving but also the State-Destroying low leakage power mode, two leakage power controlling policies for L2 cache are proposed.One is the Conserved State-Preserving and State-Destroying （C-SP＆SD） policy;the other is the Speculative State-Preserving and State-Destroying （S-SP＆SD） policy. Only one copy of a data block in cache hierarchies is in the active mode, the others are turned into the low power mode. The State-Destroying mode with least leakage power should be turned into at most as possible if no obvious performance cost occurs. Compared with the traditional leakage power controlling policies for L2 cache,C-SP＆SD policy can save the leakage power of L2 cache more efficiently without obvious performance cost;S-SP＆SD policy can save the most leakage power of L2 cache and reach the best energy efficiency of whole processor.