[FFmpeg-devel] [PATCH] Swscale YUV2RGB table generator

[Kostya]

On Fri, Feb 13, 2009 at 02:45:10PM +0200, Kostya wrote:
> On Fri, Feb 13, 2009 at 12:01:26PM +0100, Diego Biurrun wrote:
> > On Thu, Feb 12, 2009 at 10:26:21AM +0200, Kostya wrote:
> [...]
> > 
> > Here are mine from my G4, but note that libav regression tests are
> > failing anyway.  The configuration is --enable-gpl --enable-swscale:
> 
> This one looks more like the ones I got on my VIA CPU.
> 
> > diff -u -w "/tmp/ffmpeg"/tests/libav.regression.ref tests/data/lavf.regression
> > --- /tmp/ffmpeg/tests/libav.regression.ref      2009-02-07 15:06:23.000000000 +0100
> > +++ tests/data/lavf.regression  2009-02-13 11:56:12.000000000 +0100
> > @@ -24,49 +24,49 @@
> >  af78858062599fcbba049e4a02588a15 *./tests/data/b-libav.mov
> >  366879 ./tests/data/b-libav.mov
> >  ./tests/data/b-libav.mov CRC=0x45079dca
> > -913b6199765b1dcba196706574621c2f *./tests/data/b-libav.dv
> > +ce752c9e014ca11a0e4bdf1129451756 *./tests/data/b-libav.dv
> >  3600000 ./tests/data/b-libav.dv
> > -./tests/data/b-libav.dv CRC=0x7787e173
> > -8441d11ebee0465f1a672586485f3506 *./tests/data/b-libav.gxf
> > -805996 ./tests/data/b-libav.gxf
> > -./tests/data/b-libav.gxf CRC=0x9c06676e
> > +./tests/data/b-libav.dv CRC=0x645ce571
> > +0e5458da4fb47555f64cd0e3d54168b8 *./tests/data/b-libav.gxf
> > +805460 ./tests/data/b-libav.gxf
> > +./tests/data/b-libav.gxf CRC=0xd8ed3b8b
> >  66a6584f9e83e8ea3af822a3ba71fbbe *./tests/data/b-libav.nut
> >  329264 ./tests/data/b-libav.nut
> 
> Ehrm, that is strange - both has no relation to my code.
> 
> [...]
> > @@ -100,23 +100,23 @@
> >  304128 ./tests/data/b-libav-yuv422p.yuv
> >  52bb2f530036c657edc512965258d735 *./tests/data/b-libav-yuv444p.yuv
> >  304128 ./tests/data/b-libav-yuv444p.yuv
> > -345bf1aff3fa9eba908bb8ebef284346 *./tests/data/b-libav-yuyv422.yuv
> > +cd761f29085c0bf184093798d8741870 *./tests/data/b-libav-yuyv422.yuv
> >  304128 ./tests/data/b-libav-yuyv422.yuv
> >  e2d4633215bc12b933eb733cfbd9c022 *./tests/data/b-libav-yuv410p.yuv
> >  304128 ./tests/data/b-libav-yuv410p.yuv
> >  692c12328dce2b5785bd658425660296 *./tests/data/b-libav-yuv411p.yuv
> >  304128 ./tests/data/b-libav-yuv411p.yuv
> > -a0b0e086c800ee0b590c2f85bbff5e79 *./tests/data/b-libav-yuvj420p.yuv
> > +9204e83bc09e39859729a28e60751f1b *./tests/data/b-libav-yuvj420p.yuv
> >  304128 ./tests/data/b-libav-yuvj420p.yuv
> > -b365c84c3d3f3a9d6d3a7f942ac68e79 *./tests/data/b-libav-yuvj422p.yuv
> > +9c7a932ccbea3104046b1a85988a6400 *./tests/data/b-libav-yuvj422p.yuv
> >  304128 ./tests/data/b-libav-yuvj422p.yuv
> 
> That is again the part that makes me suspicious - it should not have
> any relation to yuv2rgb either.
> 
> [...]
> 
> So it looks like a bug hunt for finding out the reason.

Hmm, "make libavtest" fails on PPC even without my patch.