Fanout optimization under a submicron transistor-level delay model