The impact of cache organisation on the instruction issue rate of a superscalar processor

Much of the research on multiple-instruction-issue processor architecture assumes a perfect memory hierarchy and concentrates on increasing the instruction issue rate of the processor either through aggressive out-of-order instruction issue or through static instruction scheduling. In this paper we describe a trace driven simulation tool that we have developed to quantify the impact of the memory hierarchy on the performance of a superscalar processor that we have developed to support static instruction scheduling. We describe some initial studies performed using our simulator. As well as examining the more conventional split cache configurations, we also quantify the performance impact of using a unified cache. Finally, we examine the benefits of using two-level caches and victim caches.