Performance evaluation has become particularly overwhelming in early design tradeoff analysis. Several design decisions are made based on performance models before any prototyping is done. Usually, early design analysis is accomplished by simulation models, because building hardware prototypes of state-of-the-art microprocessors is expensive and time consuming. However, simulators are orders of magnitude slower than real hardware. Also, simulation results are artificially

sanitized

in that several unrealistic assumptions might have gone into the simulator. Performance measurements with a prototype will be more accurate; however, a prototype needs to be available. Performance measurement is also valuable after the actual product is available in order to understand the performance of the actual system under various real-world workloads and to identify modifications that could be incorporated in future designs.