Exploration of the spatial locality on emerging applications and the consequences for cache performance

The performance gap between processors and memory is increasing, making the cache hit-rate paramount for performance. Studies show room for improvement, especially in data caches. The cache effectiveness is dictated by software locality, hence the software behavior directs the cache performance. This paper presents a framework for studying spatial locality. It focuses on the characteristics of the spatial locality in terms of closeness in time and space, to get the amount of accessed sequential data and the potential for cache hits. By using the framework we gain knowledge for improving the cache performance. Our experiment consists of a program driven simulator and 11 important applications. We show a large performance potential in the data cache with up to 75% less miss rate, exploiting spatial locality. In order to investigate where potential bottlenecks are located we make a simple implementation of a scheme to exploit this spatial locality