Cache Modeling Using Spatial Locality of Reference
David S. Ferry
Dr. Jon Beck, Faculty Mentor
Every cache must implement a replacement algorithm. Traditional caches implement replacement using the temporal locality of reference of a program. In operation they attempt to keep needed data readily accessible by making the assumption that the memory most likely to be accessed is that memory which has been accessed most frequently or most recently. We propose an alternative replacement algorithm exploiting spatial locality of reference, that is, we make the assumption that data most likely to be needed is the data which is nearest to the physical address of the last data access. We will report on a simulation of our novel algorithm in comparison to the LFU and LRU algorithms in direct mapped, two-way associative, four-way associative, and fully associative caches.
Keywords: caching, memory management, algorithms
Topic(s):Computer Science
Presentation Type: Oral Paper
Session: 17-5
Location: MG 1098
Time: 10:30