On Mon, 18 Oct 1999 [EMAIL PROTECTED] wrote:
> Hi all, (A bunch of stuff on FFTs)
> Has anyone done any speed comparisons between
> - the FHT based code lame has now (which is damn fine)
> - "fastest fourier transform in the west" fftw (www.fftw.org)
> - pentium optimized fft - djbfft http://pobox.com/~djb/djbfft.html
[...]
Something which could also be very interesting to get considerable speed
ups;
Does anyone have any info about fft's with a "cut off" frequency as a
parameter ??
ie why compute (high freq) fft and psycho info if afterwards you are not
going to use/encode it; e.g. layer III hard >16Khz drop, sblimit in layer
II.
Related: I will post (today/tomorrow and for toolame) a tested patch
(will be on my www and will post to Mike) for IDCT used in
subbandfiltering i.e. IDCT(..., int sblimit)
Same thing could be used in lame (?), if sblimit is set/determined in main
program.
regards,
Patrick.
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
