[vorbis-dev] Optimizations

[Firelight Multimedia]

Ive been doing some testing lately with speed.  Ive included vorbis support
into my sound engine (www.fmod.org) for release soon..
Currently the reference source uses approximately 8% CPU in win98 on a
p3-500.  (my mp3 playback takes around 2.6% as a side note)..
This is a lot better than the previous source tree where it took around
12-13% cpu time. (sorry i cant quote the version # :) .. it is about a month
old anyway and used all doubles.

I havent gone through the source much and wont even pretend to understand
most of it, so here are some questions.
My questions are
1. Are there any algorithmic 'hacks' i can put in to speed up the playback
with acceptable quality?  I heard something about removing a dither step.
2. I am mainly a low level coder, ie i will code anything in ASM if i can
get the chance.  This should be split into 2 sub questions ->
2a.  What parts of the code are 'locked down' and can be low level
optimized?  I can (and probably will for fmod) rewrite parts of vorbis in
optimized fpu code, katmai sse and 3dnow.  I dont want to rewrite something
and have it change on me the next day :)
2b.  LPGL .. so I will resubmit these asm optimizations if people want it..
will the vorbis developers take .S files or is it too soon?

Im looking forward to releasing FMOD 3.3 with vorbis support .. It will be
the first 'user friendly' popular sound API to support it that i know of,
and should spread the word of ogg-vorbis to many thousands more developers
:)

cheers
Brett Paterson
FMOD SoundSystem
http://www.fmod.org

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.