Vorbis dev - Aug 2000 - optimization

Hi all,

I'm busy doing optimizations for vorbis. Don't have time for it, but
still
doing it :-)

Some suggestions: there are some double function names:
forward/inverse
pack/unpack
look/free_look/free_info
and maybe more.

If they were renamed to time0_forward etc. profiling would be a lot
easier.

Oh, and add -fno-inline to $PROFILE in the Makefiles (or configure).

Most expensive function in the decoder is floor0.c:inverse()
Now hacking on it...

Dag dag,

Segher

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/

> I'm busy doing optimizations for vorbis. Don't have time for it,
but still
> doing it :-)
That's the way it's supposed to be! ;-)
> Some suggestions: there are some double function names:
> forward/inverse
> pack/unpack
> look/free_look/free_info
> and maybe more.
All the backends use the same function names.  I was driving home the 
'mappingness', perhaps a bit too far.
> If they were renamed to time0_forward etc. profiling would be a lot
> easier.
true enough
> Oh, and add -fno-inline to $PROFILE in the Makefiles (or configure).
OK
>_not_ floor0.c:inverse(), but _mdct_kernel()
>Well, that's why I disable inline (at least I _thought_ I did. oops.)
Right ;-)  My MDCT drags.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/