Re: [patch] QNX-style scheduling v1.10 for Linux 2.0

Adam McKee (amckee@poboxes.com)
Mon, 15 Sep 1997 09:00:34 -0600 (CST)


> Hi!
>
> > Someone ported v1.06 or something to 2.1.x. Would someone (same person?)
> > like to port v1.10 up to 2.1.x so I can stick it in the tarball (giving
> > you full credit of course)? I think I'm pretty much done hacking on this
> > now, and so it's just basic maintenance from now on. Of course, now that
> > I've said that, someone will surely come up with a great idea to improve
> > it... :-)
>
> Hmm, it would be nice to say what is difference between v1.06 and
> v1.11...

I'll be happy to give any assistance I can to anyone who wants to port to
2.1.x.

> And a few comments:
>
> o When starting a non-time-critical CPU-intensive job that may
> take awhile to complete, you may want to put it on a run-queue > 15 to
> ensure the absolute minimum impact on interactive performance. When
> compiling a kernel, you might do 'qsched -l 16 make zlilo' -- users of
> the system would probably not even notice any slowdown!
>
> Is simply untrue. If you do not have 1gig of ram, CPU compilation will
> start swapping other tasks out (or at least it will lower number of
> disk buffers). And users are going to recognize that...

Hmm never heard the term "CPU compilation" before. Anyhow, you are right
that users could notice depending on the memory situation. On my system
(which has enough RAM) if I do 'qsched -l 16 make zlilo' then I would be
surprised if anyone noticed a slowdown.

> BTW is it still possible to use normal 'nice' to change tasks
> priorities... Looks like so.

Yes. But it's only for compatibility - use 'qsched' when possible.

-- Adam