[MPlayer-users] Option -benchmark: interpretation of results

[Reimar Döffinger]

On Thu, Jan 15, 2015 at 04:20:02PM +0100, Björn Gerhart wrote:
> Dear members,
> 
> I want to check which system load the video playback of mplayer causes.
> That's important, because video playback takes place in the background,
> while at the same time the system acts as a kind of server.
> 
> Yet unfortunately, I only can use the "vesa" xorg driver and therefore
> am using mplayer's video output "x11".

Tried fbdev directly? Not using X11 at all if possible should work
much better on such a system...

> Here is a table of the benchmarks I got with 480p versus 720p with the
> film "Big Buck Bunny" (24fps):
> 
> Resolution  VC      VO      A      Sys     Framedrop  Codec
>  854x480    49,15%   3,62%  1,28%  45,95%  0%         theora
> 1280x720    38,26%  47,71%  1,71%  12,32%  2%         ffmpeg
> 
> Now I have problems with interpreting the values. When the resolution is
> smaller, I would expect the system to take less load, which in fact is
> not the case. In my tests, the higher resolution requires less system
> load with a framedrop rate of 2% at the same time.

You should probably try videos with the same codec, encoding options
and bitrate, otherwise the comparison might not make sense.

> Could you please help me interpreting these values? Does the system load
> maybe represent the "free" system capacities and not the used one?

"Sys" means "everything that wasn't audio or video decoding".
Most of the time, that ends up being "waiting for vsync", except
that x11 doesn't have vsync support (another reason not to use it,
though fbdev isn't always better there).
Maybe your two videos are about the same bitrate and it's the
IO reading the file that is so slow?
Try using some standard tools like "top" to get a different
view on the situation.