:: Re: [maemo-leste] clutter and hildo…
Top Page
Delete this message
Reply to this message
Author: Ivaylo Dimitrov
Date:  
To: Arthur D.
CC: Merlijn Wajer, maemo-leste
Subject: Re: [maemo-leste] clutter and hildon-desktop
Hi,

On 28.02.19 г. 4:02 ч., Arthur D. wrote:
> Hello, guys.
>
> Ivaylo Dimitrov <ivo.g.dimitrov.75@???> wrote:
>> Could be, afaik Nokia did lots of work trying to squeeze the last bit
>> of performance from powervr
>
> Ivaylo, it's not about squeezing the last bit of performance.
> It's about 5 FPS vs 50 FPS.
>


No way SGX 530 rev 121 to render fullscreen with compositing on n900
with 50fps :). It is more like ~35 fps vs 12 fps but this is not really
important.

I repeat, my task back then was to make h-d work at all. It is in a no
way optimized and TBH I am not sure it is the alpha stage the best stage
to do that optimization. I know it is irritating to have h-d render with
5-15 fps, but it is not that big deal either ATM given that most of the
other devices do SW rendering.

> Merlijn Wajer <merlijn@???> wrote:
>> OK. As I indicated above, I feel the same way. But I can't help but
>> wonder if such an exercise this will bring us closer to understanding
>> the 'hang' in powervr userspace<->kernel interaction. Since it just
>> doesn't happen with old clutter and old hildon-desktop. So I am trying
>> to answer the question: what change in either clutter or hildon-desktop
>> would cause this?
>
> This part of work is already done by me a while ago. I guess I found the
> cause of the hang. The functions calls trace that leads to 2+ minutes hang
> on hildon-desktop startup is in attachment, see file "hildon-desktop".


Great!

>
> I did a simple "monkey" patch to fix the hang. If you are interested,
> please check the second attachment "hang-fix.patch". Note: it's just
> quick and dirty workaround, not a real fix. And yes, it's for cogl.
>


By looking at the trace and the patch, it seems like a bug in cogl. We
can either try to fix it or...

> If you guys really want to migrate to new clutter/cogl frameworks, I'd
> recommend to stop using deprecated functions.
>

...
Correct, there is no doubt about it. We should get rid of ClutterTexture
(and the other deprecated objects/functions) ASAP. And that may actually
fix the slow rendering.

Do you want to take on that task? I can do it, however I want to finish
connui-cellular first and that will take me some good 4-5 weeks from now.

I'll ask on IRC too, if anybody is willing to do it.

Regards,
Ivo