[cairo] scaled_font fini & directfb

[L A]

On Sun, Dec 27, 2009 at 1:58 AM, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> On Sun, 27 Dec 2009 03:32:33 +0100, Lionel Landwerlin <llandwerlin at gmail.com> wrote:
>> Le samedi 26 décembre 2009 à 16:54 -0800, L A a écrit :
>> > > Wow, somebody who is actively using the dfb backend.
>> >
>> > I guess this explains why there has not been any checkins to DFB files lately.
>>
>> ...
>>
>> > I really doubt that I am the last one though. Actually I think you
>> > might see a growing number of dfb backend users in the coming months.
>> > It seems that all the embedded companies are moving to DFB, and some
>> > projects based on Cairo are becoming popular (like webkit?).
>>
>> I'm still interested with the directfb backend :)
>> It makes sense to use directfb when you need to precisely control over
>> how much graphical memory is allocated. For example X doesn't give that
>> kind of control. For more, you can have plenty of logs when compiling
>> directfb with traces, which gives you plenty of informations about how
>> things gets done at low level (under webkit/cairo/etc...).
>>
>> Chris, if you have a better thing to use than directfb, let us know.
>
> My main issue with directfb is that it simply does not provide an
> acceleration architecture that can be used by Cairo. The best performance
> that I have been able to achieve so far is by talking directly to the GPUs
> using DRM. (That would give you even more control than directfb). In
> theory, the GL stack should come close to "bare-metal" performance. In
> practice, we are still dependent upon coding for the individual
> capabilities of hardware and we are still searching for better methods.
> --

Hi Chris,

Couple of questions, so that I understand this better:

- Could you eloborate on "simply does not provide an acceleration
architecture that can be used by Cairo"?
- What kind of an architecture are we talking about?
- What is missing in DFB?
- Why cant Cairo do transformations and scaling through DFB?
- Are there any plans for a gles (1.1 or 2.0) backend?

Thanks
Levent