ubuntu-phone team mailing list archive

[Christian Dywan <christian.dywan@xxxxxxxxxxxxx>]

I see a slight issue with that approach.

Say you're using a ton of Button's, and find that it's a bottleneck, as 
it was the case for the OSK at one point, and maybe implement a custom 
ButtonGrid component. That's fine for this particular case if you 
accept the additional maintenance overhead and it's worth it.

Now the earlier examples included MainView. That one is significantly 
harder to replace. Again it's not unheard of - *looks at Unity8 and 
Browser* - but requires you to really keep up with internal API design 
you shouldn't be dealing with and to be following design evolutions 
very closely. Otherwise you'll lose anchorToKeyboard, localizations, 
the header, and probably a few other details.

How about instead, if you really find that MainView vs Rectangle is 5s 
slower, you file a bug, and we can investigate before you start 
re-implementing everything only to find that it will be slow again once 
it's complete?

Just my 2 cents,
   Christian

Am Mi, 29. Jul, 2015 um 10:29 schrieb Ted Gould <ted@xxxxxxxxxx>:
> On Tue, 2015-07-28 at 18:45 +0200, Benjamin Zeller wrote:
> > The preloader can also create components instead of just parsing 
> > them. Which adds another
> > performance boost compared to the QML cache.
>
> So it seems like the component creation time is the biggest win here, 
> but it seems to vary on the application. In the results Zoltan posted 
> it seems that for some apps it is negligible and for some it is very 
> high. Seems likely that this is related to which components they use. 
> Perhaps we could create a similar test for each of the components so 
> that app developers would know the tradeoffs of using different 
> components.
>
> Ted