[XviD-devel] xvid_encraw update to help us tracking the hintME bug
Vincent GOMEZ
xvid-devel@xvid.org
Wed, 18 Dec 2002 16:46:08 +0100 (MET)
I had some time to look at this. And i see no bug (i don't say it's not here).
Here's my test:
- ./xvid_encraw -i cactus.pgm -m 1 -o plop_nohint.m4v
(simulates a no hinted mode)
- ./xvid_encraw -i cactus.pgm -mv 1 && \
./xvid_encraw -i cactus.pgm -mv 2 -m 1 -o plop_hinted.m4v
(First command generates the mv file with hints, the second one uses these hints to encode the 3 frames)
Results:
plop_nohint.m4v and plop_hinted.m4v are identical, a diff between the 2 files return no difference.
Consequences:
Either the hint ME runs just fine or encoder is buggy all the time.
I've done a step by step debugging and im sure the last MB is coded the same way for both hinted and not-hinted ME.
Could it be possible to have a sample from a messed up XviD content (generated with XviD stable + Hint ME) ?
Btw, the dev-api-3 bug i "real" ffmpeg complains about unfinished slices iirc, i noticed that 2 weeks ago and did not try again since this time.
@mplayer devs: do you use hinted ME the right way ? (I must ask that, no offense :-)
--
Edouard Gomez