This gives a tiny encoder speed improvement. This could have been done in 2014 after the commit 544aaa3d but it was forgotten.