Jon Noack noackjr at alumni.rice.edu
Tue Mar 15 16:30:39 PST 2005

Jon Noack wrote:
> David O'Brien wrote:
>> On Tue, Mar 15, 2005 at 01:41:55PM -0600, Jon Noack wrote:
>>> David O'Brien wrote:
>>>> On Sat, Mar 12, 2005 at 10:48:52PM +0100, Bartosz Fabianowski wrote:
>>>>>> Are you saying, all we need to do is commit this diff to make 
>>>>>> everyone's
>>>>>> environment happy?
>>>> Can you try with just -mno-sse2?  I'd like to litter the compile 
>>>> command
>>>> line as little as possible.
>>> I had this issue a while back with my Athlon-XP box 
>>> (http://lists.freebsd.org/pipermail/freebsd-current/2004-November/042127.html). 
>>> Note that I would get instant reboots with CPUTYPE?=athlon-xp, 
>>> CPUTYPE?=p3, and CPUTYPE?=p2.  However, it worked fine with 
>>> CPUTYPE?=k6-2.  I think you are right to be cautious and disable 
>>> anything that uses FP registers.
>> I remember that problem but I don't know why I don't experience it, since
>> I also use CPUTYPE=athlon-mp on my home desktop.  I suspect the problem
>> is SSE2 usage, which the Althon {X,M}P doesn't support.
> Huh.  That didn't fix it for me.  I manually merged the 
> sys/boot/i386/Makefile.inc and sys/boot/i386/boot2/Makefile changes, 
> added CPUTYPE?=athlon-xp, rebuilt world and kernel, and rebooted.  Same 
> behavior.  Oh crap -- I just noticed the sys/conf/kern.mk patch.
> Rebuilding kernel...

Well, that didn't work either.  I even rebuilt the loader with a 
hardcoded CPUTYPE=i686 for sys/boot/i386/Makefile.inc and 
sys/boot/i386/boot2/Makefile and it STILL rebooted instantly.  Is it 
possible GCC compiled with CPUTYPE?=athlon-xp produces bad code?  I 
guess my BIOS writer could've been on crack, but the machine is rock 
solid without CPUTYPE defined so I don't think it is a hardware problem.

I'm willing to debug this but I need someone's hand to hold...


More information about the freebsd-stable mailing list