On 1/27/15, Stefano Sabatini <stefa...@gmail.com> wrote: > On date Monday 2015-01-26 13:20:49 +0100, Michael Niedermayer encoded: >> On Mon, Jan 26, 2015 at 11:31:49AM +0100, Christophe Gisquet wrote: > [...] >> > To me, it looks like just a normal >> > clipping is needed. >> >> yes >> >> also on a differnt topic, i think "gamma*" should be removed from >> eq, the implementation is simply wrong and it cannot be done >> conveniently in a yuv based filter. >> we already have lut based filters for rgb space, they could easily >> do correct gamma, a "gamma" filter could be added based on that > > BTW, eq is not bitexact with mp=eq2 and differs in many parts of the > code (yes, I realized this after removing mp=eq/eq2, so this is partly > my fault). For example the LUT16 part of the code is entirely missing. > > Do we care about it? I don't, so I suggest to just keep the code and > try to improve on it rather than port the missing "features" (was a > 16bit LUT useful in that context?). > > For the future, please *ALWAYS*, *ALWAYS*, *ALWAYS*, test for > bitexactness when implementing or committing a new filter port, or > state in the log why it differs.
It IS bit exact. Michael commited options in different order than what I did in my patch. > -- > FFmpeg = Faithless and Faithless Meaningless Philosofic Epic Genius > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
