[FFmpeg-devel] [RFC]lavc/ffv1dec: Scale msb-packed output to full 16bit

[Paul B Mahol]

On 11/17/16, Michael Niedermayer <michael at niedermayer.cc> wrote:
> On Thu, Nov 17, 2016 at 09:13:55PM +0100, Carl Eugen Hoyos wrote:
>> 2016-11-17 14:49 GMT+01:00 Rostislav Pehlivanov <atomnuker at gmail.com>:
>> > On 16 November 2016 at 11:15, Carl Eugen Hoyos <cehoyos at ag.or.at>
>> > wrote:
>> >
>> >> Hi!
>> >>
>> >> Attached patch improves output for some ffv1 files imo.
>> >> Current slowdown for the existing decode-line timer is
>> >> 2%, I wonder if this can be improved through refactoring.
>> >>
>> >> Please comment, Carl Eugen
>> >>
>> >> _______________________________________________
>> >> ffmpeg-devel mailing list
>> >> ffmpeg-devel at ffmpeg.org
>> >> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>> >>
>> >>
>> > So AFAIK the encoder pushes the values to the LSBs but the decoder
>> > didn't
>> > shift them back up?
>>
>> I don't think the encoder does any shifts here but I may misunderstand.
>>
>> > I think you should add a comment explaining that happens.
>>
>> Many (older) decoders have to do this and there is nowhere a
>> comment, I really believe that this is not particularly convoluted
>> code.
>>
>> > Also 2% on a decoder doesn't sound that great,
>>
>> It's 2% in a function of a decoder.
>>
>> > did you try using an if case for the entire loop for when the
>> > values need to be shifted?
>>
>> That is what I tried to suggest with "refactoring", I suspect
>> Michael wasn't too happy about the idea.
>
> can the whole "what to put in the lsb" question be avoided by adding
> gray10 support to ffv1dec ?
> if so this might be the best solution

It is best solution and most logical one imho to support pixel format
on both input and output, expecially for lossless codec, which ffv1 should be.

>
> [...]
> --
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
>
> Old school: Use the lowest level language in which you can solve the
> problem
>             conveniently.
> New school: Use the highest level language in which the latest
> supercomputer
>             can solve the problem without the user falling asleep waiting.
>