On-chip decoupling capacitor optimization using architectural level prediction