Re: [RFC PATCH] check_preempt_tick should not compare vruntime with wall time

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

* Linus Torvalds (torvalds@linux-foundation.org) wrote:
> On Mon, Sep 13, 2010 at 10:45 AM, Mathieu Desnoyers
> <mathieu.desnoyers@efficios.com> wrote:
> >
> > FWIW, Xorg and firefox feel _much more_ responsive with the fix I propose when
> > running with a make -j10. The system is even usable with a make -j20 on my UP
> > machine, even though I can start feeling a some lag. This is probably a more
> > important, yet less scientific, result.
> 
> I'll test that myself (but in a bit - I need to go do voter
> registration and socsec update first, though - I became a US citizen
> last week).

Congratulations!

> Because yes, that's the reason I'm personally interested in your
> scheduler latency work: I think our X behavior under load is pitiful
> (I do "make -j16" on my dual-core with HT Core i5, and web browsong
> shouldn't start to lag as much as it does just because I overcommit
> the CPU a bit). So if this makes a noticeable difference, I think it's
> very important.

Please make sure to include the "smaller min_granularity" patch in your test
too, because the "check_preempt_tick" fix seems to only affect threads with
non-zero nice value. I suspect that most of the responsiveness improvement I got
is mainly from lowering the min_granularity.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com