--0ntfKIWw70PvrIHh
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Tue, Jul 11, 2006 at 07:37:17PM -0500, Peter Seebach wrote:
> In message <871wsrznpf.wl%rpaulo@fnop.net>, Rui Paulo writes:
> >At Tue, 11 Jul 2006 14:34:41 +0000 (UTC),
> >Martijn van Buul wrote:
> >> I believe there's a gcc option to use SSE2 floating
> >> point over i387 floating point, but I may be mistaken there.
> >
> >-mfpmath=3Dsse,i387
>=20
> Yes.
>=20
> And when I built a kernel with sse, I got horrible crashes and panics usi=
ng
> audio unless I went single-user.
>=20
> So there may be some quirks in how these options interact with assumptions
> made about register saving.

I think "quirk" is overly-polite.

The rule is that the kernel isn't supposed to use the floating point=20
registers or anything that overlaps them. Adding this option breaks that=20
rule; you're telling the kernel build to use floating point registers=20
after all. And things blow up. :-|

Take care,

Bill

--0ntfKIWw70PvrIHh
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)

iD8DBQFEvaHgWz+3JHUci9cRAnAIAJ41F5d95r1ngC0UVMsdu7b5qAiLfACeOnQk
47i2RdWWlKGkp7Mhjf+h2hU=
=48N0
-----END PGP SIGNATURE-----

--0ntfKIWw70PvrIHh--