Subject: Re: moving allocsys() into MI part of kernel
To: None <lukem@netbsd.org>
From: Robert V. Baron <rvb@cs.cmu.edu>
List: tech-kern
Date: 05/08/1999 23:17:32
Luke Mewburn <lukem@netbsd.org> writes:

> Hi all...
> 
> Whilst looking at which ports supported the `BUFCACHE' option (with
> intent to convert the remaining ports), I noticed that most ports

By the way, increasing the buffer space is a wonderful feature.  We
should advertise it.  Perhaps by putting a #options BUFCACHE into
the GENERIC configuration.  I was trying to see what I could do 
to make my x86 builds go faster and was going to increase BUFPAGES
but this is better.  In the table below, going to a 15% memory for
buffers reduces wall clock time by 1:30 out of 15:39.  (Sad note:
the first line is the linux 2.2.1 time.)  These measurements are
on 560X laptops with 64Meg memory 4Gig disks.

ares:L233I	514.030u 81.360s 12:09.94 81.5% 0+0k 0+0io 514653pf+0w
medon:N233I	528.430u 99.065s 15:39.80 66.7% 0+0k 12773+35452io 3756pf+0w
medon:N233I 15%	528.473u 90.731s 14:10.51 72.8% 0+0k 8682+35014io 4230pf+0w