mirror/kissfft
1
0
Fork 0
mirror of https://github.com/mborgerding/kissfft.git synced 2025-07-18 21:14:24 -04:00
Code Issues Packages Projects Releases Wiki Activity

minor doc changes

Browse Source
This commit is contained in:
Mark Borgerding
2003-12-05 00:35:25 +00:00
parent 39d2974fe3
commit 72de907478
5 changed files with 22 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
README
View File

@ -12,18 +12,18 @@ USAGE:
void * cfg = kiss_fft_alloc( nfft ,inverse_fft );
while ...
... // put kth sample in cx_buf_in_out[k].r and cx_buf_in_out[k].i
kiss_fft( cfg , cx_buf_in_out );
... // put kth sample in cx_in[k].r and cx_in[k].i
kiss_fft( cfg , cx_in , cx_out );
... // transformed
free(cfg);
Note: frequency-domain data is stored from dc to 2pi.
so cx_buf_in_out[0] is the dc bin of the FFT
and cx_buf_in_out[nfft/2] is the Nyquist bin
so cx_out[0] is the dc bin of the FFT
and cx_out[nfft/2] is the Nyquist bin (if present)
Declarations are in "kiss_fft.h", along with a brief description of the
functions you'll need to use. Code definitions are in kiss_fft.c, along
with sample usage code.
functions you'll need to use. Code definitions for 1d complex FFTs are in kiss_fft.c.
with sample usage code. For more functionality, like 2d FFTs you may need to add other source files to your project.
The code can be compiled to use float, double or 16bit short samples.
The default is float.
@ -44,8 +44,8 @@ During this process, I learned:
1. FFT_BRANDX has 500 times as many lines of code as Kiss
(and that's just the C code).
2. It took me an embarrassingly long time to get FFT_BRANDX working.
3. A simple program using FFT_BRANDX is 500K. A similar program using kiss_fft is 18k.
4. FFT_BRANDX is about 3-4 times faster than Kiss
3. A simple program using FFT_BRANDX is 522KB. A similar program using kiss_fft is 18KB.
4. FFT_BRANDX is roughly twice as fast as KISS FFT.
It is wonderful that free, highly optimized libraries like FFT_BRANDX exist.
But such libraries carry a huge burden of complexity necessary to extract every
@ -56,25 +56,22 @@ last bit of performance.
PERFORMANCE:
(on Athlon XP 2100+, with gcc 2.96, optimization O3, float data type)
Kiss performed 1000 1024-pt ffts in 110 ms of cpu time (132ms real time).
Kiss performed 1000 1024-pt ffts in 100 ms of cpu time.
For comparison, it took md5sum 160ms cputime to process the same amount of data
DO NOT:
... use Kiss if you need the Fastest Fft in The World
... use Kiss if you need the Fastest Fourier Transform in the World
... ask me to add features that will bloat the code
UNDER THE HOOD:
Kiss FFT uses a complex-only, time decimation, mixed-radix, out-of-place FFT.
No scaling is done. Optimized butterflies are used for factors 2 and 4.
Experiments with a radix 3 optimization showed no real gain over the generic
butterfly currently used for non power-2 factors.
Kiss FFT uses a time decimation, mixed-radix, out-of-place FFT.
No scaling is done. Optimized butterflies are used for factors 2,3,4, and 5.
LICENSE:
BSD, see COPYING for details. Basically, "free to use, give credit where due, no guarantees"
TODO:
*) Add sample code for parallel ffts (stereo) packed into re,im components of time sequence.
*) Add simple windowing function, e.g. Hamming : w(i)=.54-.46*cos(2pi*i/(n-1))
*) Make the fixed point scaling and bit shifts more easily configurable.
*) Document/revisit the input/output fft scaling