Subject: Re: 1.5_Beta on 3100/M76 and MVII (hare and tortoise syndrome)
To: David Brownlee <email@example.com>
From: Ken Wellsch <firstname.lastname@example.org>
Date: 10/26/2000 15:20:46
Taking your wise lead, I pulled down the syssrc for 1.3.2 and 1.4.2 (both
having available vax install sets to match) and then 1.5_BETA on my system:
Doing a ruthless editing job of the 1.3.2 GENERIC kernel I come down to
(likely I need to keep some COMPAT stuff but I chopped that too):
config netbsd root on ? type ?
backplane0 at root
cpu0 at backplane0
uba0 at backplane0
uda0 at uba? csr 0172150
mscpbus* at uda?
mtc0 at uba? csr 0174500
mscpbus* at mtc?
ra* at mscpbus? drive?
mt* at mscpbus? drive?
qe0 at uba? csr 0174440
dhu0 at uba? csr 0160440
pseudo-device loop 1
pseudo-device pty 48
Comparing 1.3.2 with a 1.4.2 hand edited GENERIC:
diff -w 1.3.2 1.4.2
< backplane0 at root
< cpu0 at backplane0
< uba0 at backplane0
> mainbus0 at root
> cpu0 at mainbus0
> uba0 at mainbus0
> rx* at mscpbus? drive?
So we switched from "backplane" to "mainbus" and a
floppy drive has been added.
And then comparing 1.4.2 with a 1.5_BETA hand-edited GENERIC:
diff -w 1.4.2 1.5.0
< cpu0 at mainbus0
< uba0 at mainbus0
> ibus0 at mainbus0
> uba0 at ibus0
< pseudo-device pty 48
> pseudo-device pty 64
So a new layer? has appeared called ibus and the pty count was upped.
One can drop the rx support as unrelated to the comparision as can any
change in the number of pty's.
David Brownlee wrote:
> On Thu, 26 Oct 2000, Ken Wellsch wrote:
> > > > What if he took a GENERIC from 1.2 or 1.3, custom stripped it down, then
> > > > carried it forward? I don't think a whole lot has changed, but I may be
> > > > out to lunch.
> > >
> > > A whole bunch - the syntax for specifying filesystems, probably
> > > some device names, options added/removed...
> > Sure in a full blown GENERIC kernel. But I'm talking about a kernel config
> > that just matches his hardware, no more. If related device names have
> > changed,
> > I'll grant you that.
> OK. Taking a few minutes to look through the file:
> Changes after 1.2
> - filesystems are now introduced with 'file-system' rather
> than options
> - ra devices attach to mscpbus rather than directly to uda
> - tapes now attach to mscpbus on mtc, not tms on tmscp
> - the 'config' line is completely different.
> Changes after 1.3
> - backplane0 has been replaced by mainbus0, which affects every
> bus attach line
> - microvaxes now have an 'ibus', to which uba0 is attached
> I may have missed some.
> > Is not the point to use a common denominator by which to judge size changes
> > and performance? Leaving in differing options is not likely to be very
> > effective comparison I think.
> I agree we should benchmark a GENERIC kernel and one specifically
> customised for the hardware and with the same set of filesystems/
> devices/network options as in the previous release.
> -- www.netbsd.org: A pmap for every occasion --