Intrinsic Data Locality of Modern Scientific Workloads

Understanding the intrinsic data locality of a workload is essential to understanding and predicting cache performance. The intrinsic data locality of a particular application or workload can be measured in a microarchitectureindependent manne,: The data resulting from these measurements ideally can be used to develop an analytic model forpredicting memory performance on different cache sizes and confrgurations. Many studies on data locality use cache hit ratios, a micmarchitecture-dependent metric, to examine locality. In this paper: we present a microarchitecturedependent and a micmarchitecture-independent characterization of the SPEC2000 workloads. We present quantitarive statistics on the diferent types of data locality (e.g.. spatial and temporal) exhibited by these workloads and we show that rhe composite intrinsic locality can be correlated to locality measured by cache hit raria.