--Signature=_Sat__12_Mar_2005_13_52_59_-0500_aVeI/MQVe3oy6Q+u
Content-Type: text/plain; charset=US-ASCII
Content-Disposition: inline
Content-Transfer-Encoding: 7bit

Hello,

> > > I run NetBSD/sparc64 on an Ultra 1 and Ultra 5 here, and really see no
> > > practical reason for running 32bit on an Ultra
> > I run a 32 bit kernel and userland on my U5 and U10 because a key
> > application has notable 64 bit bugs. The application is GCC.
> 
> I've noticed no "notable 64 bit bugs" with GCC on my U1 or U5. What bugs are
> you referring to?
Hmm, I didn't encounter any 64bit-related problems that are obviously gcc's fault but there are lots of programmers out there who expect int and void * to be the same size, so lots of applications have a tendency to barf about segfaults or bus errors (=alignment violations in this case) seemingly at random.

> I'm just stating a personal opinion that, in my case at least, I have found
> no reason not to run the machines 64-bit. One does NAT, the other is
> primarily for testing. They perform quite well enough to suit my needs.
I've read somewhere that gcc doesn't really use any v9 instructions in 32bit mode even if you tell it to, -mcpu=v9 sets only scheduling parameters in this case. No idea if that's completely true though. Then there's -mvis of course, sometimes it even does something useful ;)

have fun
Michael

--Signature=_Sat__12_Mar_2005_13_52_59_-0500_aVeI/MQVe3oy6Q+u
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (NetBSD)

iQEVAwUBQjM6i8pnzkX8Yg2nAQJBSAf/a274UjL38urFFj+sHuEfeZ32oxv16HVc
fOfHDWuzJeckh9Jt98TEe0J61L7/o6UrRQGPBjUL3nTMX41HyDiHfIKT/l9B32wm
wRIhXe8jqOcREK1c9FgupiVc8y6iMbMo0Fg+Cba9LG68ub0ggU+9OmgwPZnYMOgJ
zBQMegQ06MhisTkA/gr13lTnajbs7TzL8yEhw64QIPPJA0ctJ2bO7Xw0XDZXj/3A
Vbhze7U0ZruQhzWpjO9BR2jLJ/3DSCOkPPVT2Qx0AF4HgVVZDCxfOsZibT9xUVy2
hxUTwiw+axucMCTR0dgLLIcC0hPM7MsZQs5TyGetDz8W2dl1tHvxsQ==
=1Pu6
-----END PGP SIGNATURE-----

--Signature=_Sat__12_Mar_2005_13_52_59_-0500_aVeI/MQVe3oy6Q+u--