作者: Stanislav A. Belogolov , Jiyong Park , Jungkeun Park , Seongsoo Hong
关键词:
摘要: Embedded systems tend to use demand paging in order provide more memory applications a cost-effective manner. However, drastically degrades the performance when page fault rate is high. Prefetching has been known as common remedy for overhead. Although many prefetching mechanisms have proposed, they are either effective only specific access patterns or too straight-forward decrease an acceptable level. We propose scheduler-assisted mechanism which does not such fundamental defects. As proof of concept, our was completely implemented Linux. also conducted series experiments show its effectiveness. The experimental results showed significant improvement: number major faults and scheduling latency decreased by 30% 51%, respectively.