On Wed, May 11, 2011 at 10:20 AM, Ronald S. Bultje <[email protected]> wrote: > Hi again, > > On Wed, May 11, 2011 at 12:07 PM, Ronald S. Bultje <[email protected]> wrote: >> On Wed, May 11, 2011 at 11:59 AM, Ronald S. Bultje <[email protected]> >> wrote: >>> On Wed, May 11, 2011 at 11:26 AM, Ronald S. Bultje <[email protected]> >>> wrote: >>>> attached is a RFC for fixing SWS for 9/10bit, as output by H264. It >>>> takes endianness in account and does some other stuff. As you can see, >>>> this stuff quickly gets hideous (e.g. planarCopyWrapper is a mess), so >>>> I'd like some advice on how people think we should do this. >>>> >>>> We need this fix so we can convert 10bitBE to 10bitLE and thereby fix >>>> 10bit H264 on BE systems (e.g. PPC), which are now correct except that >>>> byte ordering for each word is inverted. >>> >>> Attached is one more that should fix fate (at least 10bit h264) on BE >>> systems, along with the swscale hackery. >> >> And one more adding a missing if. We need someone to write tests for >> these various colorspace conversions... > > And this one has fate tests added so that it doesn't break. I've > tested all YUV8/9/10/16/RGB <-> 10/9bit conversions and they all > produce images that are visibly correct, so all codepaths should be > fine. > > Ronald >
Be aware that the pixfmt.h changes break ABI. On my limited best effort basis this seems ok. _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
