Floating-point in kernel space

[Alexander Leidinger]

Quoting "Michael M. Press" <deathjestr at gmail.com> (Fri, 9 Mar 2007 18:10:25 -0500):

> > I think i've seen somewhere but i don't remember that floating point
> > arithmetic is not allowed in kernel space, if that's right, can anyone
> > please tell why ???
> > and why not not emulate the floating point in kernel space ???
> 
> That's nonsense! Floating point arithmetic is absolutely allowed in
> kernel space, it's just that user space applications don't tend to
> like their data being mangled very much.
> 
> I remember this topic coming up a short while ago on this list.
> Perhaps if there is sufficient need someone could implement a pair
> of floating-point save/restore functions to allow floating-point
> (or MMX) calculations to be used in the kernel only when needed.
> Something like the EngSaveFloatingPointState/
> EngRestoreFloatingPointState functions that Windows has.

Have a look at the FPU item on the ideas page. There you find a patch
for something like this.

Bye,
Alexander.

-- 
If there was any justice, my face would be on a bunch of crappy merchandise!

		-- Homer Simpson
		   Flaming Moe's
http://www.Leidinger.net  Alexander @ Leidinger.net: PGP ID = B0063FE7
http://www.FreeBSD.org     netchild @ FreeBSD.org  : PGP ID = 72077137