YAARC: yet another approach to further reducing the rate of conflict misses
The Journal of Supercomputing
| Hi-index | 0.00 |
Skewed-associativity is a technique that reduces the missratios of CPU caches by applying different indexing functionsto each way of an associative cache. Even though itshowed impressive hit/miss statistics, the scheme has notbeen welcomed by the industry, presumably because implementationof the original version is complex and might involveaccess-time penalties among other costs. This workpresents a simplified, easy to implement variant that we call"minimally-skewed-associativity" (MSkA). We show thatMSkA caches, for many cases, should not have penalties ineither access time or power consumption when comparedto set-associative caches of the same associativity. Hit/missstatistics were obtained by means of trace-driven simulations.Miss ratios are not as good as those for full skewing,but they are still advantageous. Minimal-skewing is thusproposed as a way to improve the hit/miss performance ofcaches, often without producing access-time delays or increasesin power consumption as other techniques do (forexample, using higher associativities).