Program Behavior Prediction Using a Statistical Metric Model

Adaptive computing systems rely on predictions of program behavior to understand and respond to the dynamically vary-ing application characteristics. This study describes an ac-curate statistical workload metric modeling scheme for pre-dicting program phases. Our evaluations demonstrate the superior performance of this predictor over existing predic-tors on a wide range of benchmarks. This prediction accu-racy lends itself to improved power-performance trade-offs when applied to dynamic power management. Categories and Subject Descriptors: C.4 [Performance of Systems]: Modeling technique