ubuntu-phone team mailing list archive

[Benjamin Zeller <benjamin.zeller@xxxxxxxxxxxxx>]

Am 28.07.2015 um 18:15 schrieb Rodney Dawes:
> On Tue, 2015-07-28 at 18:01 +0300, Zoltán Balogh wrote:
> > I know that there are security concerns about how the mapplauncherd
> > works.  After talking to the main developer (w00t on freenode) and
> > reviewing the code I have learned what the mapplaucnher is doing [7]
> >
> > If you have any experience or opinion about the topic please bring it up :)
> >
> > This idea looks very promising and piling off seconds from the app
> > starts is something we all are interest, but obviously not for the price
> > of compromising our platform's security.
> What are the real benefits of doing this? If this is to improve app
> startup speed, is there any significant and concrete data showing that a
> perceivable and significant increase in performance is made from this,
> compared against the normal execution of various types of apps, not just
> pure qml apps being run via qmlscene? Do we even have concrete profiling
> data showing where exactly the slowness is for all these various types
> of applications?
>
> This seems like a very over-engineered and complex way to work around an
> otherwise simple (albeit pervasive) problem.

The problem with the small application starting is simply the time QML
requires to be compiled first and then the time the components need to 
be created.
All this is then loaded into a internal cache for the next time a 
component is reused.

Our current solution (the QML cache) just removes the need to reparse the
components. But the first time startup experience is bad and preloading 
can scrape off another second or more.

The preloader can also create components instead of just parsing them. 
Which adds another
performance boost compared to the QML cache.

In the first mail Zoltan has linked some profiling on applauncherd vs 
qml cache vs no cache.