It's also a minor peephole optimization, but I saw bigger size wins, so I'm citing that.
m.fragment()