>>>>> "M" == Mathew Hendry <[EMAIL PROTECTED]> writes:
M> Hi all, Another day, another little optimisation. I've added
M> lookup tables to subs.c/fht() to streamline the initial
M> swapping operation for n==256 and 1024. This gives an overall
M> speedup of 1-2% on my machine (PIII 500/NT4SP5/MSVC).
It is 1-2% too, on my Celeron 464/Linux2.2.12/GCC2.95.1.
I made some hack from your code, and now I've got a more fast and dirty :)
code.
see l3psy.c and subs.c in my latest snapshot.
it merged below 3 loops into 1 loop.
- buffering and copying from savebuffer[] -> wsamp[]
- window function calculation
- first loop of FHT
I got a 2-3% fast, but the code is really dirty ....
---
Takehiro TOMINAGA // may the source be with you!
# My snapshot is available at http://www.isoternet.org/~tominaga/lame-beta/
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
