--- Richard Rauch <rkr@olib.org> wrote:
> Re.
>
http://mail-index.NetBSD.org/netbsd-help/2004/01/07/0019.html
> 
> For networking, it sounded like bochs was going to
> try to use your *real*
> network interface directly, and make the emulated
> NIC look like an NE2000.
> I didn't like that.  If I had been able to otherwise
> make bochs work for
> me, I was going to try to make it use its emulated
> serial port to do SLIP
> or maybe PPP to NetBSD.

This might work with NetBSD, but with a Winschmoze
guest? Do they do that?

VM ware does something like putting an alias on your
real network card, and then redirecting all the
traffic to that alias to the guest. That works pretty
good. I was assuming that's what was going to happen
here, but now I'm not sure. I guess I have to set up a
rule in a packet filter for this or something?
 
> CPU speed: On my AMD64, it definitely felt and acted
> a lot slower than a
> real PII 266.  (Using a Knoppix 3.1 CD as a
> reference and assuming that bochs
> would eventually have done something useful with
> Knoppix 3.1.)  It was much
> slower on an 800MHz Athlon than on the AMD64.  I
> could about believe the
> NetBSD boot message's claim about 20MHz (19.94 or
> so, actually) on the AMD64,
> though at times I was skeptical that it was even
> turning in even that much
> performance.  (^&

My Winschmoze 1998 guest felt SLOW, and I still didn't
get the mouse working. But I overwrote the "C" drive
this morning when I outsmarted myself so I get to
spend another day installing it again. I would believe
it was acting like about 15 mhz.

> I wish that I had been able to boot the NetBSD
> installation that it created.
> Then I could run a lot of familiar software and get
> a better idea of
> performance.

I just mounted the latest iso and did the most minimal
install possible with the "stock" partitioning scheme
on a 512 meg virtual disk. Worked fine, but of course,
no networking yet.
 
> pit: My guess is that if you gave it a useful tweak
> to mapping virtual
> CPU performance to real performance (I forget the
> name of the option),
> the keyboard delay would go away.  The default
> performance mapping is
> assuming something like a 200MHz Pentium, and a
> 366MHz Celeron is going
> to be much faster, so the timer for key repeating is
> going to be off.
> This would probably be a better solution in general,
> since I/O rates
> seemed to suffer terribly with "realtime" enabled.

Actually by looking at that little table in the rc
file, they look like they are set to the lower end of
a PII 400. That would be close for me, but I'll play
with it. The config.h file they want you to look at
implies that a compile time option must be enabled to
get that help with setting ips. I installed from
pkgsrc and I don't think that was included.

Anyway, I still haven't given up. This stuff is pretty
neat. It seems to have limited useability on slower
machines like I have (and I never thought the day
would come when a 1ghz Athlon [my other box] would be
slow), but on a faster one it might be viable. I think
it's still worth looking into plex86. That used to be
in pkgsrc, but it's not anymore. I'll have to figure
out why before I spend too much time with it. 

Thanks for your help.

__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus