CODEWITHSUNDEEP
cachingcpu-architectureprocessor
abeln
abeln

Reputation: 3759

Difference between cache miss rate metrics

In appendix B of Patterson & Hennessy, two different cache miss rate metrics are introduced: misses/instruction, and misses/memory-reference.

An equation relating the two is derived:

misses/instruction = misses/memory-reference * memory-references/instruction

They go on to describe the advantages of the misses/instruction metric:

The advantage of misses per instruction is that it is independent of the hardware implementation. For example, speculative processors fetch about twice as many instructions as are actually committed, which can artificially reduce the miss rate if measured as misses per memory reference rather than per instruction.

How is the above possible? Assuming the misses/memory-reference ratio is artificially lowered in speculative processors, wouldn't the equation imply that the same happens to misses/instruction?

Upvotes: 2

Views: 325

Answers (1)

aminfar
aminfar

Reputation: 2345

No, because one should use misses per committed instructions (not misses per fetched instructions, not misses per executed instructions).

Upvotes: 1

Related Questions