Things that need to be done...

1. mpg123 could pick up new sample rates suggested by the output modules (like a jack server fixed to 96kHz) and adapt to that.

Though the practical rates for MPEG audio are up to 48kHz ... but one could easily upsample.

2. Optimize float decoding

At least the 32bit float decoders should be able to work with 3DNow/SSE ... just skip the integer conversion.

3. Get SSE/SSE2 on AMD64 going.

MPlayer has that in mp3lib, which is a stripped-down mpg123 decoder.
We need that optimization, but in the non-stripped mpg123.

4. Prevent ID3v2 tags from being parsed multiple times after seek.

I need to carry a list of ID3v2 frame addresses that already have been parsed into the data structures.
Currently, this is a possible memory leak when the seek index is disabled (tag data at file beginning) or id3 data is just somewhere in the stream.
That being said, in the "normal" case, there is no leak.

5. What's about SINGLE_MIX?
Check what is _really_ happening there, make some test file...

