摘要
位置相关信息服务中访问数据涉及到复杂的空间计算,导致访问数据的延迟时间较长,而数据预取能够显著提高数据的访问速度,缩短访问数据的时间。基于LDD的预取策略如DDP考虑了数据距离,但是没有考虑数据的访问概率和更新频率及数据大小。针对以上问题提出基于价值的数据预取(CDP)策略,一些重要的数据预取因素如访问概率、更新频率、数据项大小、数据距离和有效范围等都包含在价值函数里,根据价值函数值的大小来选择被预取的数据。通过实验对比,CDP比DDP策略能更有效的提高缓存命中率。
Accessing data related to complex spatial calculations leads to a longer delay time in location dependent information services. The data prefetch can greatly improve data access speed and shorten the delay time. The prefetch strategy, based on LDD, such as DDP, considers the data distance except for the access probabilities, update rates and the size. So, the cost -based data prefetch （CDP） strategy is proposed, which takes such important prefetch factors as access probabilities, update rates, data size, data distance, valid scope, etc. into cost function. The prefetch data is selected according to the value of the cost function. The contrast experiments show that CDP is more effective than DDP strategies to improve the cache hit rate.
出处
《微处理机》
2014年第1期61-64,共4页
Microprocessors
