Workload design: selecting representative program-input pairs

Abstract

Having a representative work-load 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, it is impossible to select a huge number of benchmarks and respective input sets due to the large instruction counts per benchmark and due to limitations on the available simulation time. In this paper, we use statistical data analysis techniques such as principal components analysis (PCA) and cluster analysis to efficiently explore the workload space. Within this workload space, different input data sets for a given benchmark can be displayed, a distance can be measured between program-input pairs that gives us an idea about their mutual behavioral differences and representative input data sets can be selected for the given benchmark. This methodology is validated by showing that program-input pairs that are close to each other in this workload space indeed exhibit similar behavior. The final goal is to select a limited set of representative benchmark-input pairs that span the complete workload space. Next to workload composition, there are a number of other possible applications, namely getting insight in the impact of input data sets on program behavior and profile-guided compiler optimizations

Workload Modeling for Computer Systems Performance Evaluation

“…The most popular benchmark suite for general-purpose computing is SPEC CPU from the Systems Performance Evaluation Consortium [654]. Note that program behavior also depends on input, so representative inputs are also needed [202,204], and indeed, benchmarks specify not only the applications to run but also the reference inputs to use. Benchmarks are discussed further in Section 9.2.4.…”

Identifying "representative" workloads in designing MpSoC platforms for media processing

“…(2). In this dendrogram we can clearly distinguish between two major groups of clips: still and motion videos 1 . This kind of a coarse-grained division into two groups would have been possible to obtain just by viewing the videos on the screen.…”

“…Usually these are properties that have a strong impact on the performance of the architecture being designed. For instance, in microarchitectural design such properties would be instruction mix, branch prediction accuracy and cache miss rates [1]. As mentioned in the previous section, our hypothesis is that on the system level the performance of multimedia MpSoC architectures is largely influenced by various kinds of data-dependent variability associated with the processing of multimedia data streams.…”

“…We believe that first, the measure of dissimilarity between VCCs having identical types should be computed using Eqn. (1). These measures can then be combined in various possible ways, one of them being simply computing the sum of all the dissimilarity measures for the individual VCC types.…”

“…Recently Eeckhout et al [1] have shown that the workload design space may be very complex and therefore should be systematically explored during the construction of representative workloads. Their workload design space consists of program-input pairs that capture both, the variety of programs as well as various input data sets to those programs.…”