[FFmpeg-devel] [PATCH] Make decoding alpha optional for some codecs.
donmoir at comcast.net
Fri Sep 20 01:54:03 CEST 2013
> On Thu, 19 Sep 2013 18:50:34 -0400, Don Moir wrote:
>>I pretty much don't use ffplay for anything.
> so this patch has no effect on you whatsoever.
I could swear someone asked for comments.
> you havent even tested, but are arguing that this patch will affect
> the quality of decoding? in something you dont use ? seems pretty
> simple to check md5 of the frames before and after this patch to see
> its binary identical.
I test more than anyone else I know of except people who test for a living. I also been looking at images in extreme detail for
years. I offer a suggestion in that your use case is limited. Adding things like this just adds clutter in my opion. We don't even
know what it actually saves. All we have seen so far is something like 30% CPU. That means nothing except perhaps someway to
exaggerate. I didn't take it that way but if you take one plane and throw it away, ok it may mean some amount of less work. But it
is washed by having to read it in the first place.
> if its binary identical and speeds up decoding... what are you
> complaining about ?
o - It should be identical if I do not chose that option.
o - It won't be identical if I chose the option and alpha is just thrown away. Red value of 254 with alpha of 127 will give red
value of 254 instread of red value of 127 (over black)
o - In this case, alpha is just thrown away. Not in this case, but if you must eliminate alpha, and convert to RGB, then after much
time testing over years, then choose black as the background component or offer optional background color.
Not saying it will effect me because I would not use that option. At least it should not effect me if everything goes right :) Since
I believe this to be clutter then it just adds to the clutter. I think time would be better spent reducing that clutter instead of
introducing something like this.
More information about the ffmpeg-devel