end of proof-of-concept on imx53qsb video

[Simonas L.]

Hello Aaron,

On 19 March 2012 10:21, Aaron J. Seigo <aseigo at kde.org> wrote:

> On Sunday, March 18, 2012 21:51:37 Simonas L. wrote:
> > Dear all,
> >
> > Here is the last outcome of the proof-of-concept on KDE PA2 on Freescale
> > i.MX53 QSB:
> > http://www.youtube.com/watch?v=8kbQU6n8Zkw
>
> neat :)
>
> TA :)


> i have seen better performance on similarly (and worse) spec'd ARM devices,
> however, so i'm wondering what might be causing some of the issues seen in
> the
> video. one thing i did notice was when creating a new activity there were
> graphical glitches that look to be driver related?
>

I did not notice such glitches (namely when on-screen keyboard pops up)
when running other tests: gles_tests and Inflexion UI demo in Linaro Ubuntu
11_09, Lipstick QML of Mer. They used the same imx_dri.so driver and same
set of lib*GL*.so

>
> there are also visible artifacts on e.g. the panel shadow which is
> rendered as
> a black block or when applications are started (similar atrifact as when
> creating an activity).
>

Yes, still all other tests were probably not using those artifacts to
reveal the glitch.

One more thing I noticed: if you go 720p on that youtube video, you can see
the screen being dotted, and when the items move around or refresh, the
dotting temporarily disappears. This occurs only when GLES support is in
effect.

it's a very interesting start, but sorting out which problems are at the OS
> layer and which are due to PA itself would be very helpful to knowing
> where we
> need to focus attention (as we can improve the latter, but do less about
> the
> former)
>

Tell me how could I be of help. You could snippet a test to target
individual rendering areas, and I could launch it.
Surely if you are not getting such glitches with your end-boards, probably
no sense in wasting time here (unless you will want one running on the
Freescale's SABRE ;))


> > GLES HW accel is in effect, you can see a significant improvement, but
> the
> > responsiveness is still not real time.
>
> keep in mind that the shell itself is still using QML1 which does not have
> the
> full opengl rendering system that QML2 does. i suspect (based on playing
> with
> it :) that when we switch to QML2, we'll see a much more useful and visible
> jump in performance with GLES accel HW. one more reason to push forward to
> Qt5
>
> Will keep that in mind, thanks!


>  --

Aaron J. Seigo
>

Simonas


> _______________________________________________
> Active mailing list
> Active at kde.org
> https://mail.kde.org/mailman/listinfo/active
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/active/attachments/20120320/b7b9fa7a/attachment.html>