Re: gcc -m68060 option in kernel Makefiles

[Ignatios Souvatzis <is%netbsd.org@localhost>]

On Sat, Jan 17, 2009 at 12:02:44PM +0100, Frank Wille wrote:
> Ignatios Souvatzis wrote:
> 
> > Hm, what does Amiga have? Back then, I knew what I did - which I don't 
> > do any longer by heart, and I'm 400 km away from my 68060 book.
> > 
> > I seem to recall that -m68060 is fine as long as you don't do floating
> > point stuff (same as 68040, btw!) or MMU, and we don't do FP stuff
> > in kernel C code, and gcc doesn't do MMU stuff, so it wouldn't make
> > a difference.
> > 
> > The FPU stuff is explicit limited precision arithmetic - contrary to
> > setting the precision and then do the generic operations, which is 
> > what you have to do with the 68881 or 68882.
> > 
> > Do I recall correctly?
> 
> IIRC...
> 
> - The 64/32 integer division (DIV?.L Dn,Dr:Dq) has to be emulated.
> - The 32*32->64 bit integer multiplication (MUL?.L) has to be emulated.

Yes, and -m68060 avoids that. What I wanted to say - -m68060 does not
*add* any instructions we use in the kernel.

        -is