Cache Timing¶
Average Access Time¶
\[Average Mem Access Time = t_{hit} + r_{miss} * penalty_{miss} \]
Improve Hit Time¶
- The smaller the cache, the better the hit time.
- The lower the associativity, the better the hit time.
However, these will greatly increase the miss rate.
Reduce Miss Rate¶
- The larger the block size, the lower the miss rate. Having a very big block makes movement of data too expensive to be worth though. Leads to unnecessary data movement during miss.
- The block size is determined more so by the data access pattern than optimising for the miss rate.
- Increasing cache size to X2 reduces miss rate by root 2 (empirical rule)
- Direct-mapped cache of size N has same miss rate as 2-way associative cache of size N/2