On Mon, 2 Dec 2002, Jared D. McNeill wrote:

> > > For example, consider that you reserve at least 30% of memory for exec
> > > and files. Having done so, you can NEVER achieve an anonmax of 95% --
> > > at most, 70% can be used for anons.
> >
> > most cases but not never.
> >
> > if on 1GB machine i will just run 100kB program doing scientific
> > computation and needing 950MB of RAM it will get it with anonmax 95%.
>
> No it won't, Wojtek. If that were the case, 'filemin' and 'execmin' would be
> completely useless. Think about it.

Actually, it can. If a big process causes everything else to swap out,
then when it exits, you've got less then 'execmin' resident. Same
thing for when a lot of processes exit, as when an X user logs out.

I second raising execmin to 10, as that helps a lot for me several
different setups. I was running a tiny NFS server (1.6J)  with only
16mb for a little while, but it took an hour to boot up with execmin
at 5, no problem at 10. Computers with more memory run more stuff: my
workstation/Xserver/Gateway/Fileserver with 192mb (1.6_STABLE) swaps
out every window that's not in the foreground at 5, but behaves nicely
at 10, and my kid's laptop that I just set-up with 1.6_STABLE (64mb)
also runs Gnome much better with execmin at 10.

Frederick