#8 Dropping b-frames causes instability or loss of sync

closed-fixed
6
2005-04-09
2005-03-22
Nicholas Hardy
No

Currently any frame is open for dropping. Sometimes when
certain frames are dropped it confuses xvid. Previously
this caused on assertion in decoder.c.

I've removed the assertion, it's still happening and seems
to have major consequences for sync.

I've started looking at changing the whole frame dropping
code to be a bit better, should fix most of this problem in
the process.

Discussion

  • Nicholas Hardy
    Nicholas Hardy
    2005-03-23

    • priority: 5 --> 6
     
  • Logged In: YES
    user_id=691941

    Hm, I was under impression that the only type of frames
    that are possible to be skipped are b-frames. Isn't it so.
    Can it really drop regular P-frames?

     
  • Nicholas Hardy
    Nicholas Hardy
    2005-03-25

    Logged In: YES
    user_id=657618

    Hmm...maybe one of these days I should find out more about
    how other players use mpeg4. On the other hand I'm having too
    much fun just hacking the code randomly :)

    My understanding is that there is no basic technical reason why
    any frame cannot be dropped. In the current version even
    key-frames get dropped sometimes. However I'd guess that the
    visual impact of dropping b-frames would be much lower since as
    I understand/guess they contain the differences in between
    other frames and have no/little effect on the decoding of other I
    and P frames.

    I'll look at just dropping b-frames in the future, but I think some
    changes I'm working on right now are improving the situation
    anyway.

     
  • Nicholas Hardy
    Nicholas Hardy
    2005-04-02

    Logged In: YES
    user_id=657618

    I think I may getting closer to a solution.

    I've found that if you delete TML from the palm and reinstall it,
    then run a movie the sync appears to be better. I think the key
    may be changing/accessing the preferences for the first time
    after install. Could also be the first reset or something like that, I
    can't reproduce the bug reliably enough right now.

    Could anyone confirm this behaviour on their samples, and - if
    possible - help in narrowing down the bug.

     
  • Nicholas Hardy
    Nicholas Hardy
    2005-04-09

    • status: open --> closed-fixed
     
  • Nicholas Hardy
    Nicholas Hardy
    2005-04-09

    Logged In: YES
    user_id=657618

    Fixed in r81