My meeting notes for 2023-03-20

[Sharaf Zaman]

Hi Dmitry!

> 1) Just refactor the dependent-resource fetching/marking and don’t do the checks in the delegate. That would remove the whole metadata
> fetching line from the delegate.

So, we just create a map to mark broken preset as “broken” and somehow keep them
in sync with the database (I’ll have to investigate how)?


PS: I’m not very well-versed with the dependent-resource system (probably
evident from the fact that I removed code in delegate rendering :p).

Dmitry Kazakov <dimula73 at gmail.com> writes:

> Hi, Sharaf!
>
> I have just checked the code in VTune. It seems like the line you mentioned takes about 10% time while scrolling. Not that it is too much, but
> still comparable with 15% spent on the thumbnail scaling :)
>
> I guess we have two solutions here:
>
> 1) Just refactor the dependent-resource fetching/marking and don’t do the checks in the delegate. That would remove the whole metadata
> fetching line from the delegate.
> 2) Make sure that KoResource::metadata() is always synchronized with the database. I guess it should be somewhat synchronized, though I’m
> not very sure :)
>
> On Mon, Mar 20, 2023 at 1:08 PM Dmitry Kazakov <dimula73 at gmail.com> wrote:
>
>  HI, Sharaf!
>
>  Do you mean that you see this metadata line in the profiler when rendering/scrolling the preset chooser?
>
>  I can profile this testcase under VTune and share results if needed.
>
>  On Mon, Mar 20, 2023 at 12:43 PM Sharaf Zaman <shzam at sdf.org> wrote:
>
>  Hello!
>
>  I hope everybody had a good weekend and has been having a good Monday :)
>
>  To give an update on what I did last week: Firstly I merged most of my standing
>  merge requests, I merged my MR to improve rendering of Brush Presets, although I
>  had to make some modifications because they were breaking broken presets. So, it
>  is next in line for some optimizations.
>
>  I merged my MR for easier switching of resource locations and I fixed a bug with
>  cursor icon being stuck as a “hand” (bug 456183).
>
>  And finally I looked into caching prepared SQL statements that we use. I first
>  asked about it on phabricator (<https://phabricator.kde.org/T15245#288770>) and
>  was told nobody has looked into it before. I had noticed that QSqlQuery::prepare
>  being shown high up in the profiler, so I thought it might be a good idea to
>  optimize this away, but before I do a big search and replace, I want to make
>  sure with something more localized to see if it actually can improve performance
>  in Krita or not.
>
>  Kind Regards,
>  Sharaf Zaman
>  <https://www.sh-zam.com>
>
>  –
>  Dmitry Kazakov