[MPlayer-users] feature request

[D Richard Felker III]

On Wed, Oct 29, 2003 at 10:55:13AM -0800, Sycotic Smith wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> Wie geht's!
> 
> I have a small request that came to mind when I was thinking about mplayer in generalized terms for my own uses. It can be done several different ways, you can decide.
> 
> Thoughts:
> 1) Mplayer binaries compiled with runtime cpu-detection should be
> able to 'store' the value found on first-run in the config file. ie:
> mplayer installed from rpm runs the first time and finds an
> 'athlon-xp' cpu, writes line to config file 'cpu=athlon-xp' and
> after next run reads that in to speed up detection/startup.

Hmm? Do you notice the 1ms delay at startup to detect cpu? IMO reading
an extra line from disk takes longer. :)

> or,
> 
> 2) Commandline option given to binary install with runtime
> cpu-detection turned on, stl '-cpu athlon-xp' and then uses commands
> for that cpu.

It's not detecting the cpu type that makes runtime cpu detection
slower. It's the fact that you have extra conditionals in the inner
loops (either actual conditionals or calls through function pointers)
rather than explicit calls to the code optimized for your cpu.

> Are either feasible options, or both? Just brainstorming, maybe I
> need a braincoat?

:)))

Rich