Recency-based TLB preloading

Abstract

Caching and other latency tolerating techniques have been quite successful in maintaining high memory system performance for general purpose processors. However, TLB misses have become a serious bottleneck as working sets are growing beyond the capacity of TLBs.