Subject: Re: tuning for small memory machines
To: Simon Burge <simonb@wasabisystems.com>
From: Brian Chase <vaxzilla@jarai.org>
List: tech-kern
Date: 11/29/2002 13:28:42
On Wed, 27 Nov 2002, Simon Burge wrote:

> I'm running with vm.anonmax=95 now.
>
> Now, the question is should we automatically tune vm.anonmax up high
> for low-memory situations?  And what would qualify as a "low-memory"
> situation?

FWIW, out of curiosity, I decided to tweak the vm.anonmax value on my
16MB MicroVAX 3100 (avail memory 12212KB) running the 1.6 release.
I was hoping this might improve my compilation performance a bit.

For my test, I compiled two identical kernels.  I used the `time'
command to time the compiles.

  With the default vm.anonmax=80, the `make netbsd' time is as follows:
  16479.544u 996.689s 4:59:27.65 97.2%    0+0k 141+9374io 754pf+76w

  And by setting vm.anonmax=95, the `make netbsd' time is:
  16487.698u 991.133s 4:59:23.60 97.3%    0+0k 154+9285io 723pf+78w

I was astonished with how close the results were, and obviously a bit
disappointed to not see any gains.  Maybe there are none to be had
with my whopping 12MB of available memory? :-)

-brian.