Subject: Re: 1.5_Beta on 3100/M76 and MVII (hare and tortoise syndrome)
To: Anders Magnusson <ragge@ludd.luth.se>
From: NetBSD Bob <nbsdbob@weedcon1.cropsci.ncsu.edu>
List: port-vax
Date: 10/27/2000 12:00:58
> > on the MVII, even with a stripped to the bare bones 1.5 kernel. Ethernet
> > speeds, for example were 33% faster on the 1.3.2 with generic kernel,
> >
> This is true. It's even faster than that. Something has happened in the
> IP stack that makes it much slower on slow machines between 1.3 and 1.5;
> I posted about this on tech-kern a couple of months ago. I have checked
> all obvious details like device drivers and at last it was only the IP
> stack left.
The IP stack would make that much difference? Interesting.
I recompiled a stripped minimal kernel on both 1.5B1 and 1.3.2,
and same thing. Also, the 1.3.2 kernel was 412k while the 1.5B1
was 650k. Same config file. Any thoughts as to what has balooned
in the code, even with a stripped config? Maybe I ought to compare
the sizes of the binary objects and see if there is a pattern to
the bloat. That might point to where things are happening.
I picked up another MVII, today, so I will have two machines to run this
up on and compare bits.
This really has my curiosity piqued.
Bob