5.4 Measuring and Improving Cache Performance
Cache association
Cache line的映射方式主要有三种：directed mapping、full associative、set associative。
可以通过灵活的映射方式来reducing miss rate。
Directed mapping：
每个memory block只能存在cache中一个cache line。
每次加载一个新的memory block，都要把某个cache line替换掉，miss rate较大。
Full associative：
每个memory block可以存放在cache中的所有的cache line。
在确定cache是否hit的时候，需要较多的硬件资源；一般适用于cache line number较少的cache。
Set-associative 组相连
一般是n-way set associative（n路组相连）。
每个set有n个cache line组成，每个memory block可以放在某个set的任何一个cache line中。
Way代表一个memory block可以放在几个cache line中。
从另一个角度来说，三种mapping方式都是set -associative，只不过way不同
Locating block in set associative cache
在set associative cache中，地址中的index field是用来索引哪个set的，到底是set中的哪个cache line还要通过比对tag来决定。对比tag的过程是同时进行的——如果是4 way，那么需要四个比较器。
下图中是4way cache，共有256个set。
Replacing block strategy
一般是LRU（least recently used）
Reducing the Miss Penalty Using Multilevel Caches
In particular, a two-level cache structure allows the primary cache to focus on minimizing hit time to yield a shorter clock cycle or fewer pipeline stages, while allowing the secondary cache to focus on miss rate to reduce the penalty of long memory access times.
the primary cache of a multilevel cache is often smaller. Furthermore, the primary cache may use a smaller block size, to go with the smaller cache size and also to reduce the miss penalty.
In comparison, the secondary cache will be much larger than in a single-level cache, since the access time of the secondary cache is less critical.With a larger total size, the secondary cache may use a larger block size than appropriate with a single-level cache. It often uses higher associativity than the primary cache given the focus of reducing miss rates.
参考：
1. 计算机组成与设计 第五版 ARM版