[Mplayer-users] dirty ideas to speedup libMP3

Christoph Lampert lampert at math.chalmers.se
Mon Jul 23 12:45:39 CEST 2001


>> Oh... boooring! Anybody can do that! I'd rather prefer a real-time version. 
>
>i didn't get it. the CPU is too slow to play it in real time, so you'd prefer
>it to play it... in real time??

Yes, that's exactly my point :-)

Let me explain in detail: My CPU is too slow to play the video, so mplayer 
used  -framedrop (or even -hardframedrop) to skip parts of the video, either
by not copying them to video memory or simply by not decode them. 
The picture does not run as smoothly, but it runs. 

That's not reasonable with sound (as discussed earlier) because skipping
sound is much worse to the human senses than skipping picture. 

But you may also have noticed, that low-bitrate video needs less CPU to be 
decoded than high-bitrate and the same with audio: low-bitrate MP3 decode faster 
than high-bitrate. 

So I am looking for an algorithm to: Either reduce the bitrate of the audio 
stream _before_ decoding it (e.g. by decoding only every second byte, yes, that 
will not work, but just as a metaphor), or decode only parts of the audio stream 
(e.g. when doing inverse cosine transformation "forget" about the highest 
frequencies) Both would be faster with reduced audio quality, but not result in
skipping sound, which is horrible to listen to.

I'm not a MP3-expert, but something like that should be possible... if not
1) which depend heavily on file format, than at least 2)... Whom can I ask about 
mp3lib-decoding routines?

gruel

-- 
Dipl. math. Christoph Lampert (complex analysis, integral formulae)
Email: gruel at gmx.de                |     Email: lampert at math.chalmers.se


_______________________________________________
Mplayer-users mailing list
Mplayer-users at lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/mplayer-users



More information about the MPlayer-users mailing list