[Ffmpeg-devel] Re: [PATCH] QTRLE encoder

Clemens Fruhwirth clemens
Thu Feb 15 17:07:04 CET 2007


At Thu, 15 Feb 2007 01:23:20 +0100,
Baptiste Coudurier <baptiste.coudurier at smartjog.com> wrote:
> 
> Clemens Fruhwirth wrote:
> > Thank you Michael for your very detailed reply. I read all your
> > suggestions and I agree with all of them.
> > 
> > I only infinitesimally favour to see this patch merged in contrast to
> > an unmerged patch. That's why I won't allocate a personal time slice
> > to play the merging game.
> > 
> > Thank you nonetheless as I'm sure this will be a great help for
> > anybody who has this objective.
> 
> Here is an updated patch. Issues should have been adressed, Michael if
> you see more simplification/optimization, I'll do it. I'll add
> regression tests also, if wanted.

I'm glad that someone picked this up and invested time to make this
patch acceptable to the ffmpeg team.

But I have to object against removal of RGB555 mode as this is
certainly the most useful mode. Screencasts with a color depth of
24-bit are a bit too much of luxury. I can understand that the
maintainers do not like to see inefficient code merged (as put_pixel),
but there is no excuse for eliminating working features that are
implemented inefficiently but _implemented_.

For all that do not nod up until this point: This point of view can
also be argued with performance being your first imperative. You need
an infinite amount of time to encode an RGB555 animation movie with
ffmpeg without a qtrle encoder supporting RGB555 mode. This is the
once-in-a-lifetime opportunity to realize the biggest performance
increase in this area by actually adding this feature.

But except for this single point of critique, thanks for your work&time!
-- 
Fruhwirth Clemens - http://clemens.endorphin.org 
for robots: sp4mtrap at endorphin.org






More information about the ffmpeg-devel mailing list