Re[2]: [Gnash-commit] gnash configure.ac ChangeLog doc/C/sources.xml

[Udo Giacomozzi]

Hello Rob,

Monday, June 4, 2007, 7:01:07 PM, you wrote:
RS>   I am not convinced this correctly identified the problem. You backed
RS> out everything I changed and it made no changes, so I fail to see the
RS> point.

Not everything. I could not test your Makefile/configure patches and
that's why I was asking for your help. I did various builds before and
after that commit and *all* showed that performance decreased
dramatically after the commit. It would be great if at least somebody
could confirm the symptoms by simply building both versions and
running profiler.swf.
http://www.gnashdev.org/wiki/index.php/Profiling exists for that
purpose.

I'm currently loosing lots of time with this problem - time I'd prefer
to invest in other Gnash development tasks. I still /use/ a Gnash
checkout from May 09th because current Gnash CVS version is not
usable for me.


RS> I am sorry you are obviously upset that AGG is not currently the
RS> default for people that build Gnash from source.

Yes, I expected you would say that. No, it's not the point that I did
write the AGG backend. It's a fundamental thing. It would make no
difference for me if the topic hypothetically was "The FFMPEG backend
plays all sounds correctly, but we default to GStreamer - even if it
can't play some sounds - because it uses less CPU.". So, again, I'm
*not* upset. Instead, I fear we will get bug reports related to OpenGL
rendering quality. Sure, people will complain about lacking
performance with other renderers, but won't file a bug report for
that.


RS> This is *not a big deal*, let's not blow this out of proportion.

Okay.


>> BTW, is AGG really so much slower when playing YouTube so that one
>> notices the difference?

RS>   Yes.

Interesting! Then please tell me about it. YouTube videos play very
smooth on my laptop with about 20% CPU usage last time I checked.
Please tell me more about your experiences.



RS> I totally appreciate the
RS> energy and time you have put into the AGG renderer. My wanting to use GL
RS> as the default for source builds is not meant to slight AGG at all.

Again, I have no problem with non-AGG renderers! I'm really objective
in this regard...


RS> I'm just trying to reduce the incoming stupid bug reports.

We want the same thing, but have different predictions.


RS>   Some of us *are* focused on the release, so it's normal we have
RS> differing priorities.

So I conclude that performance is a priority, so *please* let's have
a look at that commit. I expect that it's a stupid little thing we
will find *together* in maybe an hour so we can remove that brake
block without much effort and boost Gnash rendering performance with
*all* rendering backends. Does that sound good?

Udo