Did a little more work.
Original mp2 code took 120 seconds to encode a 40 second file.
Removed all the pow()s and did a couple of other bits, and it's now about
55 seconds. (all on my p166).

According to ccett (http://www.ccett.fr/audio/c7_audio.htm) realtime mp2
encoding should be very possible on a p100. [there's also an article on
this page about mp2 coding worth reading]

Also, an amiga 060/66 can encode an mp2 at a speed of 2 times realtime (ie
2 minutes to encode a 1minute sample). [although this encoder is heavily
laced with tuned ASM afaik)
Based on mp3 encoding speeds, my p166 is about 5 times faster than an
060/66, so my computer should be able to encode an mp2 in about .5
realtime.

The CCETT paper (mentioned above) highlights the fact that the encoder
side of mpeg audio isn't defined strictly.  You can adapt the models
however you want, so long as it produces as a valid stream.  They tweaked
the spychoacoustic model to be less cpu intense, but still made a better
sounding audio stream.

things to think about if you're sick of looking at mp2 code
mike
(ps i've gone back to the mp2 code partly to see if I can really grok it,
and then move onto the mp3 code)

Reply via email to