On Sat, 15 Sep 2001, Daniel Haglund wrote:
> I just got linuxbios up and running. :) I am using a DoC with a K7SEM and a
> Duron 750MHz. Things are working rather well but I have a performance
> problem.
Well, here's the problem. At AMD's request I have had to remove the K7
support from LinuxBIOS. The slowness you see is a result of that removal
of support.
AMD's Open Source Relations people tell me I can not currently release the
5 lines of C that would make your K7SEM work well. They are trying to
figure out if they will EVER allow us to release this code. I am trying to
put together a list of companies to get in touch with AMD to try to get
this resolved. They are committed to getting us an answer within 30 days.
They are trying as hard as they can, but they can not make any guarantees.
There are three possible ways out of this.
1) AMD lets us release that code. For the record, I have spoken to a CPU
designer at AMD about this. He tells me that this material should never
have been NDA in the first place. He was actually surprised that
it was under NDA. It is essentially NDA by accident.
2) Some bright fellow reverse engineers a BIOS (it would be VERY early in
startup ...) and tells us how to turn the chip on. Denis Dowling, where
are you now :-)
3) We all give up on AMD and use Intel only. In which case
all of us who spent money on K7 hardware are screwed, and all the
companies that put man-hours into K7/LinuxBIOS are screwed too. This
would be very unfortunate. I do like these K7s. But at this point
we have had to suspend all our K7 projects. I am really glad I didn't
buy the real big AMD cluster I was planning on getting ...
> Does this have anything to do with L2 cache by any chance? Naturally I had
> to comment out the #SORRY line in freebios/src/cpu/k7/cpufixup.c.
It's not an L2 cache thing. You might be able to figure it out by looking
at early startup code in any AMD bios. Just look for stuff that is not
documented in any publicly available document, and take it from there.
ron
