> Date: Mon, 14 Aug 1995 16:19:07 +0100 (BST)
> From: David Brownlee <D.K.Brownlee@city.ac.uk>

> 	PS: Why is the kernel compiled with -msoft-float? (Ok, I know the
> 	answer is 'so it will work without a FPU' - but I wondered if it
> 	would be possible to automatically disable that flag if FPU_EMULATE
> 	is not defined in the config file? - or is -msoft-float required
> 	for the kernel even on machines with an FPU?)

Actually, this is paranoia the prevents accidental FPU instructions.

Regardless of whether there is an FPU, the kernel must NEVER use
floating-point instructions that modify any of the floating point
registers, because the kernel does not maintain its own set of FPU
registers.  The onlf FPU instructions in the whole kernel should be
the ones used to save/restore the FPU state on behalf of a process
during process switch, and the FPU probe/initialization code.

Gordon