[Libav-user] [FFmpeg-user] ffmpeg burst traffic problem

[Zach Swena]

In general hardware decoders have a problem with the bursty nature of
FFmpeg's UDP.  I have fixed this by buffering to disk and sending it with
my own program.  ISO131808-1 does not specify the maximum jitter allowable
in packets, but I have found up to 500 micro seconds to be acceptable for a
4mbps mux.  The problem is bursty UDP datagrams cause underflow and
overflow of the input buffer in hardware decoders.

The pacing solution I used for my own program was to make an event loop
thread that only ever performs "short" operations.  This thread is simply a
loop which checks a timer with nanosecond resolution.  If it is less then
1/3 the way through the period, I process other operations, but as it nears
the time to transmit, it just functions as a simple wait loop.  We really
need someone familiar with the UDP code to rework the transmit pacing.


Zach

On Fri, Oct 16, 2015 at 9:21 AM, julian at jusst.de <julian at jusst.de> wrote:

> > I had a patch once that would "slow down" udp output if you'd like to
> > try it...what is the problem though, are you unable to receive it on
> > the receiving side?
>
> Do you happen to still have that patch around? I just ran into a problem
> where ffmpegs udp output is too bursty for the used receiver. Would be
> interesting to try your patch.
>
> -Julian
>
>
>
> --
> View this message in context:
> http://libav-users.943685.n4.nabble.com/Libav-user-ffmpeg-burst-traffic-problem-tp4658964p4661598.html
> Sent from the libav-users mailing list archive at Nabble.com.
> _______________________________________________
> Libav-user mailing list
> Libav-user at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/libav-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ffmpeg.org/pipermail/libav-user/attachments/20151018/89bab7da/attachment.html>