| Computers Forum Index » Computer - Graphics (Algorithms) » Looking for high performance FFT/IFFT codes in open... |
|
Page 1 of 1 |
|
| Author |
Message |
| Johnson... |
 Posted: Tue Sep 29, 2009 1:51 am |
|
|
|
Guest
|
I am looking for an high performance open-source C/C++ codes for FFT and
IFFT computation. I hope its performance isn't much worse than the Intel
IPP library.
Could anybody please provide some information?
Thanks. |
|
|
| Back to top |
|
|
|
| Nicolas Bonneel... |
| Posted: Tue Sep 29, 2009 2:25 am |
|
|
|
Guest
|
Johnson a écrit :
Quote: I am looking for an high performance open-source C/C++ codes for FFT and
IFFT computation. I hope its performance isn't much worse than the Intel
IPP library.
Could anybody please provide some information?
Thanks.
There is fftw which is very efficient. There are benchmark at:
http://www.fftw.org/speed/
Otherwise, a gpu implementation is available and can be faster for large
sizes, using Cuda. It is called cuFFT.
Cheers
--
Nicolas Bonneel
http://www-sop.inria.fr/reves/Nicolas.Bonneel/ |
|
|
| Back to top |
|
|
|
| Nicolas Bonneel... |
| Posted: Tue Sep 29, 2009 2:32 am |
|
|
|
Guest
|
Nicolas Bonneel a écrit :
Quote: Johnson a écrit :
I am looking for an high performance open-source C/C++ codes for FFT
and IFFT computation. I hope its performance isn't much worse than the
Intel IPP library.
Could anybody please provide some information?
Thanks.
There is fftw which is very efficient. There are benchmark at:
http://www.fftw.org/speed/
Otherwise, a gpu implementation is available and can be faster for large
sizes, using Cuda. It is called cuFFT.
(if you merge the benchmark FFTw with the benchmark cuFFT:
http://www.science.uwaterloo.ca/~hmerz/CUDA_benchFFT/
you can see that, although FFTw can be slower than IPP library, cuFFT
can be faster for large matrix sizes ) |
|
|
| Back to top |
|
|
|
| Johnson... |
| Posted: Wed Sep 30, 2009 5:16 am |
|
|
|
Guest
|
Nicolas Bonneel wrote:
Quote: Johnson a écrit :
I am looking for an high performance open-source C/C++ codes for FFT
and IFFT computation. I hope its performance isn't much worse than the
Intel IPP library.
Could anybody please provide some information?
Thanks.
There is fftw which is very efficient. There are benchmark at:
http://www.fftw.org/speed/
Otherwise, a gpu implementation is available and can be faster for large
sizes, using Cuda. It is called cuFFT.
Cheers
Thank you very much, Nicolas. That is exactly what I am looking for!
BTW, do you have the source codes of CuFFT and FFTW? Otherwise do you
know where I should look for the source codes?
Johnson |
|
|
| Back to top |
|
|
|
| John Tsiombikas... |
| Posted: Mon Oct 26, 2009 5:16 am |
|
|
|
Guest
|
On 2009-09-30, Johnson <gpsabove at (no spam) hotmail.com> wrote:
Quote: Nicolas Bonneel wrote:
There is fftw which is very efficient. There are benchmark at:
http://www.fftw.org/speed/
Otherwise, a gpu implementation is available and can be faster for large
sizes, using Cuda. It is called cuFFT.
Cheers
Thank you very much, Nicolas. That is exactly what I am looking for!
BTW, do you have the source codes of CuFFT and FFTW? Otherwise do you
know where I should look for the source codes?
Nicolas was kind enough to give you a url for fftw, you might as well
take the time to look through it before you ask where to download the
source code...
And if he hadn't, google is your friend.
--
John Tsiombikas (Nuclear / Mindlapse)
http://nuclear.sdf-eu.org/ |
|
|
| Back to top |
|
|
|
|
