On 26/04/2019 13:43, Thorsten Behrens wrote:
> Right - IIRC that was very trivial hairlines there, not much value in
> buffering that -
=)
>> * should we not disable the SystemDependentDataBuffer ie.
>> remove:
>>
>> if(maEntries.empty() && maTimer)
>> maTimer->Stop();
>>
>> from there we can stop ourselves in the timeout if necessary.
>
> Not sure I get the question - surely disabling the timer helps to get
> scheduler load down? As this code is already inside implTimeoutHdl().
I rather suspect that adding and removing a timer a very large number
of times has a potentially significant cost. Whereas adding a timer once
- and (when it happens a second later) - noticing there is nothing to do
and removing it has a very small cost once per second. Also makes
logging & debugging scheduler related issues much nicer.
> Just a bit concerned we'd flip-flop between optimizing for different
> cases here - some perf test rig would be ideal I guess..
If we get the low hanging fruit I guess we win and leave the detail to
the future =)
ATB,
Michael.
--
michael.me...@collabora.com <><, GM Collabora Productivity
Hangout: mejme...@gmail.com, Skype: mmeeks
(M) +44 7795 666 147 - timezone usually UK / Europe
_______________________________________________
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/libreoffice
