Hi all
a number of things...
1) lamer 2.1f is out. This has an improved filter_subband() routine. About
50% faster than the old one. Thanks go to James Droppo.
Reference: K. Konstantinides, "Fast subband filtering in mpeg audio
coding" IEEE Signal Processing Letters, 1(2), Feb 1994, pp 26-28
[NB the dodgy resampling routine is still in there. I'll improve or remove
when i actually get some quality programming time]
2) encoding is dependant upon filetype!!
My default file is track18.aif. It takes 113 seconds to encode with 2.1f
If I convert the file to track18.raw (just a raw format pcm file) it takes
only 107 seconds to encode. Interesting and silly. If someone susses out
why this is so, tell me. [I don't have time to investigate at the moment,
still too many assignments to do]
3) If you're looking for something to do..
Go to http://noel.feld.cvut.cz/~pollak/WWW/icassp96/html/ic96s212.htm
and grab the paper titled "A high performance software implementation of
mpeg audio encoder" by Kumar and Zubair.
This paper has a window_subband() speedup, the same filter_subband()
speedup as mentioned above, and most importantly it has a very interesting
bit_allocation speedup.
This last speedup is for layerII code, but as someone has mentioned
previously, why does bit_allocation need to be iterative? This speed up
does bit allocation in log(n) rather than n time , by using a heap rather
than iteration.
I suppose if someone was amazingly keen, they could adapt it for layerIII.
that's all for now
mike
