Re[2]: [Gnash-dev] what is 'klash' ?

[Udo Giacomozzi]

Hello Tomas,

Tuesday, October 31, 2006, 8:13:17 PM, you wrote:
>>   Actually the version of AGG in all the distributions is a potential
>> show stopper, as we don't want to require anyone to build AGG from
>> source to build Gnash. So we're still brainstorming. One big advantage
>> OpenGL has is simply that was how the last release went out, so that's
>> what people expect.

Agree. This could be a problem.


TG> I'm voting for the OpenGL as default. When playing some swf, like
TG> gangsta_rap_se.swf (google for it), AGG looses the sync/tempo, and i'm on a
TG> core 2 duo! Maybe it's just a bug... But accelerated OpenGL is still my
TG> personal preference over AGG.

That movie takes up 90% even with the Adobe Flash player under Windows
even at reduced window sizes. Looks like it (the movie) is badly
implemented. 

The OpenGL renderer is of course faster when you have the hardware for
it. The only slowdown could be the tesselator but rendering
performance will certainly be much better and it's logical since all
the hard work is done in the gfx card, not in the cpu you're
monitoring.

Also, there is yet much to optimize in the AGG renderer, since the
primary goal was to get a full featured renderer, even if it may never
reach the hw-accel OpenGL performance, though.

The most difficult movies for AGG are probably those with *very*
complex shapes and those with many very large shapes. Shapes with
large radial gradient areas might be a problem too, as the code
contains a sunoptimal workaround.

I'd like to know the difference for Ninjai movies.


The optimal way would be to use AGG for standard gfx cards and OpenGL
for for hardware accelerated ones. Since that can't (currently) be
detected at runtime it's better IMHO to go the more universal way for
reasons I already described.

So, my opinion is to not decide the renderer in base of the individual
speed but what will satisfy more people out of the box.

However, as Rob already pointed out, the AGG version mess could be
really a show stopper. So, if the compatibility layer approach really
can't help here and there's no other solution I agree that AGG may be
a problem and OpenGL might be better. For all other reasons I'd use a
solution that is not so much hardware dependant (much faster with hw,
much slower without)...

Udo