[FFmpeg-devel] Interested in applying SoC @ ffmpeg

Måns Rullgård mans
Fri Mar 20 20:47:50 CET 2009


Janez Urevc <janez at ratatuj.tv> writes:

> 2009/3/20 M?ns Rullg?rd <mans at mansr.com>
>
>> Robert Swain <robert.swain at gmail.com> writes:
>>
>> > On 20/3/09 17:16, Michael Niedermayer wrote:
>> >> On Fri, Mar 20, 2009 at 05:51:36PM +0100, Janez Urevc wrote:
>> >>>> I added your name, but note that a simple patch for pngdec.c that does
>> the
>> >>>> rgb48->rtb32 conversion is
>> >>>> 1. unlikely to reach svn
>> >>>> 2. not enough as a qualification task.
>> >>>>
>> >>>> I can't say anything about adding rgb48 support to libswscale, though.
>> >>>
>> >>> How much will this one take?
>> >>
>> >> rgb48 input shouldnt be too hard if you can make sense of the swscale
>> >> code, parts of it are in need of a cleanup ...
>> >
>> > The swscale code is quite complex anyway, before considering that it is
>> > rather difficult to read in many places because of macros giving
>> > indirections. Or as M?ns put it, the code makes his eyes bleed.
>> >
>> > I just wanted to note that diving into this code won't be the easiest
>> > thing but don't give up, persevere and one should get there. As Michael
>> > says, submit your code early and persist with it and you'll get much
>> > more out of it and learn much more about our standards and how we
>> operate.
>>
>> IMO, before anyone touches libswscale again, it should be thoroughly
>> cleaned up.  I will NOT do this because it is riddled with x86
>> assembler which I don't understand, nor do I understand how it is
>> supposed to work.  If someone does clean it up, chances are I'll add
>> some ARM optimisations, but not before.
>
> So is this apropriate qualification task or not? I can start to work on it
> if you think it is ok to to do that or I can also take something else if it
> is better so.

Of the last week's commits to libswscale, roughly half have broken
compilation in one or more configurations, the rest being fixes for
said breakage.  This says something about the quality of the code.

-- 
M?ns Rullg?rd
mans at mansr.com




More information about the ffmpeg-devel mailing list