Jonathan Stone wrote:
> 
> >Hi,
> >
[...]
> >the /133 is about *ten times* slower than the /240... The /240 is
> >humming along at about 35-40 kKey/sec. while the /133 crawls at about 4
> >kKey/sec...
> 
> The obvious question is, "How fast does the client ru on a native
> Ultrix 5000/133"?

Ok, time to go down to the basement and pick another /133 from the pile
and give it a try :-) Based on raw CPU speed difference I'd expect the
/133 to run at ~3/4 the speed of a /240 (???)

> 
> >Is there something in the Ultrix emulation that causes this dramatic
> >slowdown? Both machines are idle most of the time...
> 
> Simon Burge ran some tests with CPU-intensive, statically-linked
> Ultrix binaries, on two 5000/260s, one running NetBSD and one running
> Ultrix.  The Ultrix binaries consistently ran faster on NetBSD.

Don't ever mention /260 while I'm around :-) Last week I've been told of
the demise of several *dozen* /260s that have been brutally destroyed
(hammered to pieces) because no one knew what to do with them
:-(((((((((((((((((((

> 
> I'm not awaere of anything in the Ultrix emulation which would cause
> such slowdowns. But you could try running a ktrace() on another
> copy of the client to see.
> 

Yup, it's time to get out all those nifty tools and dig around...

> The difference here is that your RC5 client is a very long-lived
> process; it may be running into scheduler bugs. NetBSD 1.3.3 inherited
> some scheduler bugs from 4.4bsd. (iirc, the bug only really impacts on
> machines with clockrates noticeably higher than 60Hz, so the pmax is a
> candidate for trouble).  what version of netbsd are you running?

The kernel is

   NetBSD decstation.magrathea.frankfurt.netsurf.de 1.3I NetBSD 1.3I
(GENERIC)    #0: Tue Feb  9 22:39:33 PST 1999       
                           ^^^ (forgot to setup the right timezone...)

One thing I forgot to mention in my original post is the fact that the
/133 is running diskless off of a FreeBSD server...

By the way: I get some rather confusing (or funny?) output from 'ps' for
two other 'long-lived' processes, namely 'portmap' and 'cron'

USER   PID %CPU %MEM   VSZ  RSS   TT  STAT STARTED       TIME COMMAND
[...]
root   78  0.0 -508.6  68 -166680 ??  Is   14Feb99    1:28.66 portmap -l
[...]
root  125  0.0 -151.3 144  -49592 ??  Is   14Feb99    0:29.65 cron

The machine has been up for about 20 days now, mainly chewing on RC5 key
blocks...

thanks,
Marc