Optimize floating point initialization
Lowl3v3l in #osdev on 2017-02-17 points out the floating point unit is initialized in two steps but could be done in one step, saving about 3 cycles, and points to the pastebin at http://pastebin.com/4vtxQpxZ with an example of how it could be done alternatively. It does the same but is somewhat faster due to parallel execution.