[MPlayer-users] mencoder "cluster"?

Richi Plana mymplayer at richip.dhs.org
Fri Sep 28 02:49:09 CEST 2007


On Thu, 2007-09-27 at 16:46 -0700, RC wrote:
> On Thu, 27 Sep 2007 17:32:20 -0600
> Richi Plana <mymplayer at richip.dhs.org> wrote:
> 
> > Does anyone know how DVD::Rip's
> > <http://www.exit1.org/dvdrip/doc/cluster.cipp> clustering
> > functionality work and how something similar (or the same) could be
> > done using mencoder?
> 
> I imagine it just splits the video into chunks, and encodes
> them independently on different machines.
> 
> You could do the same easily using something like avisplit to split the
> file into however many chunks you have computers, and using ssh to copy
> them over, and execute the mencoder command.  

I imagined that was what dvdrip must be doing, but is there a way to get
the same functionality of avisplit from mencoder? You see, the way
dvdrip clustering works, it relies on the source multimedia data to be
accessible on the cluster slaves. This is usually done via NFS, but
other virtual filesystems should work. It doesn't split the file so I
imagine transcode has a way to select "blocks" from it's input (so long
as it's random accessible ... is that even the right phrase?).

Can one break down an input video into sections and tell mencoder to
work on a certain section only? What would these "blocks" be? Frames?

> Though, to get full quality, you really need to run the 1st pass on a
> single system (rather fast if using 'turbo'), then very carefully split
> the file on what should be I-frame boundaries, and the divx2pass.log
> file at the exact same frames as the file, then do only the 2nd pass on
> seperate machines.

Also, dvdrip clustering processes the audio as one whole block on the
machine with fastest access to the input source (usually the one which
has the file on its harddisk). I can't see why the first pass in a
2-pass encoding scheme couldn't be done that way.

> Personally, I don't think either is worth the hassle, unless you have a
> large number of rather slow systems, and also absolutely require fast
> turnaround of each individual video for some strange reason.

I do have a couple of fast machines, and believe me ... after getting
used to dvdrip clustering, it's kind of hard going back to 1-machine
processing. It takes so much longer and, personally, I'd rather make use
of the otherwise spare cycles.
--

Richi Plana




More information about the MPlayer-users mailing list