[FFmpeg-devel] [PATCH 1/2] avformat/avformat: Introduced `AVInputFormat.read_timestamp2` to fix keyframe seeking for formats that rely on `read_timestamp` for seeking

Darren Mo fumoboy007 at me.com
Sun Sep 8 01:52:01 EEST 2019


I think I understand some of what you are saying. You are saying that after the bisection, continuously step backwards + read forward while doubling the step size until the keyframe is found?

Some questions:
- What is the difference between the lowercase k and the uppercase K in your diagram? And to make sure I didn’t misunderstand, v is the result of the bisection and A/B/C indicate the steps?
- Why after finding the keyframe in step C, is there an extra -->?
- Re: “there is code somewhere in git to do something similar already”, do you have an idea of what I should search for to find this code?

Thanks!

> On Sep 3, 2019, at 4:39 AM, Michael Niedermayer <michael at niedermayer.cc> wrote:
> 
> stepwise searching for the keyframe in forward direction while taking steps
> backward
> ......................k.K.................v-start
>                                     A--->
>                            B------->
>            C-------->k-K-->
> 
> there is code somewhere in git to do something similar already
> this should be faster than always searching for keyframes during bisection as
> long as the initial step is reasonable in relation to the keyframe distance
> 
> also as both approaches are quite similar one could even choose which
> way to do it depending on the expected cost of seeks and bandwidth and
> if some statistics on the keyframe distance are known and or the
> search direction
> 
> thx
> 
> [...]
> 
> -- 
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
> 
> There will always be a question for which you do not know the correct answer.
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".



More information about the ffmpeg-devel mailing list