Patches item #1055258, was opened at 2004-10-27 08:52 Message generated for change (Comment added) made by eighthave You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=1055258...
Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: puredata Group: None Status: Open Resolution: None Priority: 5 Submitted By: Tim Blechmann (timblech) Assigned to: Miller Puckette (millerpuckette) Summary: fftw support
Initial Comment: this patch provides the possibility to use the fftw library instead of the builtin mayer fft ... fftw is at least twice as fast ...
it is a compile-time feature, so pd can be compiled without the fftw ...
----------------------------------------------------------------------
Comment By: Hans-Christoph Steiner (eighthave)
Date: 2006-01-17 08:41
Message: Logged In: YES user_id=27104
http://www.fftw.org/install/windows.html
According to that website, it sounds like FFTW builds fine with SSE on MinGW. So that should work well, since everything else in CVS builds with MinGW too.
----------------------------------------------------------------------
Comment By: Ed Kelly (edkelly) Date: 2006-01-17 08:10
Message: Logged In: YES user_id=1322859
Have you checked out the Windows binaries of FFTW3 at http://www.fftw.org/install/windows.html ? Maybe it's a hassle, but there are build files for VC6 and precompiled binaries...perhaps the problems mentioned on this page with SIMD instructions would be addressed by a more recent MSVC, but I only have MSVC6 so I can't check this...
Ed
----------------------------------------------------------------------
Comment By: Hans-Christoph Steiner (eighthave) Date: 2005-12-02 00:48
Message: Logged In: YES user_id=27104
This patch is safe since it does not enable any modified code unless you set the HAVE_LIBFFTW3F define. It would be nice to have the option to use FFTW already in the source.
----------------------------------------------------------------------
Comment By: Guenter Geiger (ggeiger) Date: 2005-05-18 03:47
Message: Logged In: YES user_id=430561
There is another fast FFT package called oura, i haven't benchmarked it personally, but it seems to be quite fast. Its a "one source file" FFT, so it could be used as a replacement for the meyer FFT (except, it computes RIRIRI instead of RRRIIII)
http://momonga.t.u-tokyo.ac.jp/~ooura/fft.html
No ALTIVEC or SSE support in it I fear.
Guenter
----------------------------------------------------------------------
Comment By: christopher clepper (cclepper) Date: 2005-05-18 01:15
Message: Logged In: YES user_id=663212
I have used fftw on OSX and the Altivec version is many, many times faster than the Pd FFT. The installation was not too difficult, although there are tons of compile time options and the defaults are not usually the most efficient ones. For the port of partconv~ I just statically linked it for the binary, which added a few hundred Kbytes but it seems well worth it.
Overall it looks like the FFT lib to adopt for any cross-platform app that requires efficient implementations on multiple architectures. The code compares quite well to commercial libs and applications.
----------------------------------------------------------------------
Comment By: Miller Puckette (millerpuckette) Date: 2005-05-18 00:54
Message: Logged In: YES user_id=313747
I don't know what to do about this one. I don't use FFTW because it requires "configure" and I don't have a windows machine with CYGWIN, so wouldn't ever be able to test this. Perhaps it's for use with linux only? Rumor has it fftw is much faster than meyer, but I've never wanted to go to the hassle of learning how to install it.
----------------------------------------------------------------------
You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=478072&aid=1055258...
-
SourceForge.net