[MPlayer-dev-eng] [PATCH] ao_sdl buffering change

[Reimar Döffinger]

Hi,

>>this is something I've been playing around with a bit. It doesn't make
>>too much sense except from a theoretical standpoint, but if you have
>>time please look at it and comment, test etc. anyway.
>>Currently ao_sdl uses a ringbuffer consisting of 8 seperate buffers, for
>>which I can't see any reason at all. In addition it isn't thread safe,
>>which it should be as SDL might use a different thread for playback of
>>the sound.
>>This implementation uses a type of ringbuffer that should provide enough
>>thread-safety for this special case while not requiring synchronization.
>>I'm not sure if this is more likely to increase or decrease performance,
>>I doubt there will be much change. Feel free to test if you want ;-)
> 
> Looks like as the cpu usage ,displayed in the status bar, is at least twice as 
> high with audio only files here. Otherwise it seems to work fine.

Hm, that's not good. Although the reason probably is related to the fact 
that the mplayer.c loop is spinning much too fast in the audio-only case 
(makes gmplayer use 100% CPU for playing MP3 files on my Athlon 800), 
but I fear touching mplayer.c...
Thanks for testing!

Greetings,
Reimar Döffinger