AN IMPROVED FUZZY REPLACEMENT ALGORITHM FOR CACHE MEMORIES

A fuzzy replacement algorithm (FRA) for cache memories has been proposed (Hossain et al., 1991). In this paper we present a modified fuzzy replacement algorithm (MFRA) that improves the performance of the cache memory manyfold over the FRA. The following notable changes have been made to the FRA: the number of inputs has been reduced from three to two, the number of fuzzy linguistic variables and the dimensions of the universes of discourse have been increased, and a larger knowledge base that consists of 34 new rules is employed. The performance of the MFRA, as shown in the results, is comparable with the performance of traditional replacement algorithms, such as least recently used (LRU) and first in first out (FIFO), under real work load conditions. The interesting feature of the MFRA is that the hardware (Hossain and Manzoul, 1993) proposed for the FRA can readily be used to implement the MFRA.