Evaluating the Impact of Cache Capacity on AMAT under Sequential and Shuffled Access Patterns

Authors

  • Ebtihal Al-Ajeeli Ali Al-Mrabet University of Azzawia, Azzawia, Libya Faculty of Science Department of Computer Science, Author
  • Fathi Hamhoum University of Azzawia, Azzawia, Libya Faculty of Science Department of Computer Science Author
  • Rawad Masoud Salman University of Azzawia, Azzawia, Libya Faculty of Science Department of Computer Science Author

DOI:

https://doi.org/10.5281/echzd143

Keywords:

Cache Size, LRU, Sequential Access, Shuffled Access, AMAT.

Abstract

Cache memory plays an important role in improving computer system performance by reducing memory access time. This study investigates the effect of cache size on memory access performance using a C++-based cache simulator. Four cache capacities (4 KB, 8 KB, 16 KB, and 32 KB) were evaluated using the Least Recently Used (LRU) replacement policy and two memory access patterns: Sequential and Shuffled. The experiments used a 32 KB working set, a cache line size of 64 bytes, and 100,000 memory accesses for each configuration. Performance was evaluated using cache hits, cache misses, hit rate, miss rate, Average Memory Access Time (AMAT), and relative improvement between consecutive cache sizes. The results showed that the 4 KB, 8 KB, and 16 KB caches achieved an AMAT of 101 ns under the Sequential pattern, whereas the 32 KB cache reduced it to 1.51 ns. Under the Shuffled pattern, AMAT decreased from 88.61 ns for the 4 KB cache to 1.51 ns for the 32 KB cache. These results indicate that cache capacity can significantly affect memory access performance when the cache becomes large enough to accommodate the working set. The study provides a practical comparison of cache sizes under consistent simulation conditions.

References

مجلة القلم المنير للعلوم الانسانية والتطبيقية

Downloads

Published

2026-09-02