Experimental shite.

Experimental shite.

The kernel doesn't get passed CPU flags from the functions called here.
Assuming the kernel doesn't seek backwards from the exheader pointer,
the CPU state should already have been set by AM, which means loader
can override what was previously set.

That being the case, if I am correct, setting the CPU speed from
loader should technically persist to application launch. This needs
testing, obviously, but I believe granular CPU speed control is
possible from here.