[FFmpeg-devel] rectification filter

Daniel Oberhoff danieloberhoff at gmail.com
Tue Jul 29 07:41:10 CEST 2014

First off thanks for taking the time to review.

Am 29.07.2014 um 01:16 schrieb Carl Eugen Hoyos <cehoyos at ag.or.at>:

> Daniel Oberhoff <danieloberhoff <at> gmail.com> writes:
>> + * This program is free software; you can redistribute it and/or modify
>> + * it under the terms of the GNU General Public License
> Your patch is missing a configure hunk that 
> makes sure that the filter is only compiled 
> if --enable-gpl was specified.
> Search for *filter_deps="gpl" for an example.

That notice is left over from the original rectification filter from which I started, but since I effectively rewrote every single line and ended up reusing only the algorithm I am going to re-license the code under lgpl.

>> +// todo: only include what's needed for the given vector width
>> +#include <emmintrin.h>
>> +#include <immintrin.h>
>> +#include <pmmintrin.h>
>> +#include <x86intrin.h>
> I suspect this is not acceptable.
> There are multiple reasons iirc, among them compiler 
> compatibility.
> I suggest you submit a C only version first, yasm 
> optimizations (or theoretically inline asm but 
> this would also cause some opposition) can be 
> added later.

Ok, I think I can live with that for now. Sad to see that there is still no portable way to exploit vector intrinsics in c. Could be an interesting exercise
to write a yasm kernel though.

> See tools/patcheck for a tool that (also) tells 
> you about a few style nits, since it is your 
> file, you could choose to ignore them, but it 
> makes everybody's life (slightly) easier if 
> you follow our general style.

I was not going to ignore them and will do that.

More information about the ffmpeg-devel mailing list