I suspected that -j wouldn't do anything on a single file. For three large (6.5GB in total) files I'm getting good performance, about 1.6x of what GNU Grep does, best case.