The below two steps are implemented in order to kill the BCMDPC process:
1: Rebooting the switch when you have a maintenance window is the simplest way to bring down the high CPU utilization by bcmDPC process.
2: clearing the forwarding database seems fixing this issue as well. The command is clear fdb?. The switch will have to re-learn all forwarding information which should be unnoticeable, but may cause short slowdown in traffic (depending on the number of FDB entries).