Addressing workload variability in architectural simulations

The inherent variability of multithreaded commercial workloads can lead to incorrect results in architectural simulation studies. Random perturbations and statistical methods help computer architects draw valid conclusions.