L2missrate

L2 miss rate is a measure used in computer architecture to indicate how often an access to the L2 cache misses. It is the fraction of L2 cache accesses that do not find the requested data in the L2 cache and must be fetched from a higher level of memory hierarchy. It is commonly expressed as a percentage or, in performance analysis, as misses per thousand instructions (MPKI).

Calculation and interpretation. The L2 miss rate is computed as misses_L2 divided by accesses_L2. For example, 2 million L2 misses out of 10 million L2 cache accesses yield a 20% L2 miss rate. When normalized to program work, it can be reported as MPKI by calculating (misses_L2 / instructions) × 1000. A lower L2 miss rate generally correlates with better performance, while a higher rate indicates more frequent stalls waiting for data from slower memory levels.

Factors and relationships. The L2 miss rate depends on cache size, associativity, line size, and replacement policy, as well as program characteristics such as working set size, locality, and access patterns. It is influenced by prior L1 behavior; a high L1 miss rate often correlates with a higher L2 miss rate, though the relationship is not fixed. The overall impact on performance also depends on the miss penalty and the ability to overlap memory accesses with computation.

Measurement and use. Analysts measure L2 miss rate using hardware performance counters, performance modeling, or simulators. It is a key metric in performance tuning and architectural design, used to assess cache effectiveness, guide optimizations for data locality, and compare processor or memory subsystem configurations. Caution is warranted when comparing miss rates across different systems, as counting methods and cache hierarchies can vary.