>> Right. But is igvn.remove_dead_node(n) better/different than pushing on >> igvn worklist? > > It's more explicit and saves another round of igvn processing of the dead node (and other nodes that > become dead as well). Ok. I'll make that change. Roland.