Date: Sat, 17 Nov 2007 07:35:12 +1100 From: Peter Jeremy <peterjeremy@optushome.com.au> To: Bruce Evans <brde@optusnet.com.au> Cc: stable@freebsd.org, bde@freebsd.org, Pete French <petefrench@ticketswitch.com> Subject: Re: Float problen running i386 inary on amd64 Message-ID: <20071116203512.GD20992@server.vk2pj.dyndns.org> In-Reply-To: <20071117025943.P64582@delplex.bde.org> References: <20071115062002.GK89746@server.vk2pj.dyndns.org> <E1IseEn-000315-Uy@dilbert.ticketswitch.com> <20071115184615.GL20992@server.vk2pj.dyndns.org> <20071117025943.P64582@delplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--5I6of5zJg18YgZEa
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Sat, Nov 17, 2007 at 04:53:22AM +1100, Bruce Evans wrote:
>Behaviour like this should be expected on i386 but not on amd64. It
>gives the well-known property of the sin() function, that sin(x) !=3D sin(=
x)
>for almost all x (!). It happens because expressions _may_ be evaluated
>in extra precision (this is perfectly standard), so identical expressions=
=20
>may sometimes be evaluated in different precisions even, or especially,
>if they are on the same line.
Thank you for your detailed analysis. Hwever, I believe you missed
the critical point (I may have removed too much reference to the
actual problem that Pete French saw): I can take a program that was
statically compiled on FreeBSD/i386, run it in legacy (i386) mode on
FreeBSD-6.3/amd64 and get different results.
Another (admittedly contrived) example:
jashank% uname -a =
=20
FreeBSD jashank.vk2pj.dyndns.org 6.1-STABLE FreeBSD 6.1-STABLE #15: Wed Aug=
2 18:35:57 EST 2006 root@jashank.vk2pj.dyndns.org:/usr/obj/usr/src/sy=
s/jashank i386
jashank% cat y.c
#include <stdio.h>
double one =3D 1.0;
double three =3D 3.0;
double third =3D 1.0/3.0;
int main(int argc, char **argv)
{
if (one/three =3D=3D third)
puts("Equal");
else
puts("NOT Equal");
return (0);
}
jashank% cc -O2 -fno-strict-aliasing -pipe -march=3Dathlon y.c -static -o=
y
jashank% ./y
Equal
jashank% /sbin/sha256 y
SHA256 (y) =3D d44fe8c4c4b4beab6125ba603f2a34fa4d0280ff04d697e22594debf9efc=
9a1a
jashank%=20
turion% uname -a
FreeBSD turion.vk2pj.dyndns.org 6.2-STABLE FreeBSD 6.2-STABLE #30: Tue Jul =
31 20:29:49 EST 2007 root@turion.vk2pj.dyndns.org:/usr/obj/usr/src/sys/=
turion amd64
turion% scp -p jashank:y . =20
y 100% 146KB 145.9KB/s 00:00=
=20
turion% /sbin/sha256 y
SHA256 (y) =3D d44fe8c4c4b4beab6125ba603f2a34fa4d0280ff04d697e22594debf9efc=
9a1a
turion% ./y
NOT Equal
turion%=20
This is identical code being executed in supposedly equivalent
environments giving different results.
I believe the fix is to initialise the FPU using __INITIAL_NPXCW__ in
ia32_setregs(), though I'm not sure how difficult this is in reality.
--=20
Peter Jeremy
Please excuse any delays as the result of my ISP's inability to implement
an MTA that is either RFC2821-compliant or matches their claimed behaviour.
--5I6of5zJg18YgZEa
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)
iD8DBQFHPf7//opHv/APuIcRAjUpAKDD7ALpNxKehry9KyiIscoNUPcTPACfTKnL
Y+cceitfaXLAOg/jSS53/8M=
=0/6f
-----END PGP SIGNATURE-----
--5I6of5zJg18YgZEa--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20071116203512.GD20992>