본문 바로가기
  • Home

Adaptive Set-Division Replacement Policy for Increasing Cache Efficiency

  • Journal of Knowledge Information Technology and Systems
  • Abbr : JKITS
  • 2014, 9(5), pp.589-598
  • Publisher : Korea Knowledge Information Technology Society
  • Research Area : Interdisciplinary Studies > Interdisciplinary Research
  • Published : October 31, 2014

Youngil Cho 1 Lee Sang Jeong 2

1수원대학교
2순천향대학교

Accredited

ABSTRACT

Modern processors have large on-chip caches to mitigate off-chip memory latency. The Least Recently Used(LRU) replacement policy represents the cache blocks in a set as LRU stack. This policy picks the LRU block as the replacement candidate. In most condition, it can accomplish the work well, but for some memory-intensive workload it can not provide a long enough access history for a given cache size and associativity. Also in some workload, the majority of lines go through the cache space without making any sense. Cache performance can be improved if a long range of access history can be held so that some period of records can contribute to cache hits. In this paper, we propose an Adaptive Set-division Replacement Policy(ASRP) for effective cache management. In ASRP policy, the set of the last-level cache is divided into several subsets, only one subset is activated when replacement happens and the replacement area is limited to this subset using LRU policy. The cache misses are counted when a certain subset is activated. When the miss-count exceeds a threshold, the next subset is activated. So the threshold can indirectly decide the range of access history through the last-level cache. We use a sample method to dynamically determine the threshold for different workloads and different run-time phases of a certain workload. The experiment results show that ASRP reduces the average MPKI of the baseline 1MB 16-way L3 cache by 4.6%.

Citation status

* References for papers published after 2023 are currently being built.