How Input Data Sets Change Program Behaviour

Having a representative workload of the target domain of a microprocessor is extremely important throughout its design. The composition of a workload involves two issues: (i) which benchmarks to select and (ii) which input data sets to select per benchmark. Unfortunately, we are unable to select a huge number of benchmarks and respective input sets due to limitations on the available simulation time. In this paper, we use principal components analysis (PCA) to efficiently explore the workload space. Within this workload space, different input data sets for a given benchmark can be displayed and representative input data sets can be selected for the given benchmark. The final goal is to select a limited set of representative benchmark-input tuples that span the complete workload space.