[Gnash-dev] Re: DirectFB

[Udo Giacomozzi]

Hello Rob,

Sunday, October 14, 2007, 6:26:34 AM, you wrote:
RS> I got a weird email with no sender or subject, but it appeared to look
RS> like a message that you were looking at a DirectFB and maybe a Xara
RS> backend for Gnash ?

We discussed this a few days ago in IRC but no, I didn't send you a
e-mail on this regard.

Regarding Xara I had some mail exchange about a year ago with Charles
Moir:

> Our rendering engine is called CDraw (sometimes called XaraDraw) and, at
> the moment, is the one component that we have not made GPL. It's not
> unlike Cairo in many ways. However we do plan to GPL this within the
> near future. Subscribe to our announce mail list I would suggest to hear
> news about this.
>
> As to whether it's suitable for animation is another matter. I've always
> assumed that Flash was optimized for animation from day one, and makes
> quite some compromises on quality (such as flattening, anti-aliasing and
> line stroking) in order to achieve the speed. We are not animation
> orientated, and have made no such quality compromises or optimisations
> for animation, so I can't say how we'll compare against Flash. My
> expectation is that it should certainly be fast enough for most typical
> Flash style animations.

...

> OK. If you're a commercial company I should tell you we have a much
> faster assembler optimised version of XaraDraw that's not being open
> sourced. Although thinking about it you are probably not running x86 cpu
> on embedded devices so this is probably of less interest.


Reading some more recent posts in the XaraXtreme-dev mailing it seems
that they themselves are looking for an alternative to CDraw (mainly
considering Cairo and AGG) since CDraw apparently will remain closed
source:
http://www.xaraxtreme.org/maillists/archive/dev/dev_022007/msg00117.html

So I don't think it makes sense to build a Xara backend. Probably it
wouldn't be faster than AGG either. Xara (CDraw) most probably doesn't
offer a Flash-style rasterizer so we need to use our triangulator,
which will be the bottleneck.

A Cairo backend makes more sense for it's hw acceleration
possibilities. Dunno about (dis)advantages compared to OpenGL.

DirectFB makes sense, but is a new GUI, not a new backend, and should
be fairly easy to implement using AGG.

I don't think we will find any faster software renderer than AGG. We
still need a full featured hardware renderer. Probably there's not
much missing in the OpenGL backend but AFAIK nobody is working on
that. Also, the triangulator needs some rework (as others mentioned).

Udo