Scale-Model Simulation

Computer architects extensively use simulation to steer future processor research and development. Simulating large-scale multicore processors is extremely time-consuming sometimes impossible because of infrastructure limitations. This paper proposes scale-model simulation, a novel methodology predict system performance. Scale-model first constructs simulates scale model the target with reduced core count shared resources. Target performance then predicted through machine-learning (ML) based extrapolation. predicts 32-core on single-core an average error 8.0% 15.8% for homogeneous heterogeneous workloads, respectively, while yielding 28× speedup.