[MEncoder-users] Encodes for slow laptop

[Diogo Franco]

Em Qua, 2008-06-11 às 09:16 -0700, Stella-Terra Clemens escreveu:
> resolutions, in 5 different bitrates.  Then, I'm going to create another
> shell script (though I'm having a little trouble with this) that will run
> each file in mplayer using the -vo null -nosound -noskip -benchmark and
> parsing out the results in second and storing them in a CSV file which I can
> then import into a spreadsheat and graph.
There will be no speed difference between libavcodec and xvid, except
for quality and ENCODING time because they're two different encoders for
the same specification. Only libavcodec is used for decoding by mplayer
and xvid doesn't implement a decoder so the difference between them is
IMO not worth testing.

You should also pick up a video that works and test some 'vo's and see
which one is the fastest. You can do it with '-hardframedrop -benchmark'
and see on the status line which one drops less frames, or just the one
that uses less cpu time. Remember to use '-lavdopts skiploopfilter=all'
to decode the h.264 videos, it will gain a lot of speed.

Some VOs to try:
xv
gl
gl:force-pbo
gl2
gl2:noglfinish

Before trying the gl 'vo's, check the output of "glxinfo | grep 'direct
rendering'". If it says "No", then install accelerated drivers for you
video card. If there aren't any, then don't bother testing them.

For xv, take a look at the output of "xvinfo | grep Adaptor" and test
each line on the vo. For example, with my NV18, that command outputs:
  Adaptor #0: "NV17 Video Texture"
  Adaptor #1: "NV05 Video Blitter"

Then you use '-vo xv:adaptor=NUMBER' and see which one is the fastest.
In my case, 'xv:adaptor=1' is faster than the default 'adaptor=0'. Also,
for me 'gl2:noglfinish' is the only one faster than 'xv:adaptor=1'.