fftw on FC5?

Tim Fenn fenn at stanford.edu
Fri Apr 28 18:43:42 UTC 2006


On Wed, Apr 26, 2006 at 06:29:51PM -0700, Globe Trotter wrote:
> 
> --- Tim Fenn <fenn at stanford.edu> wrote:
> 
> > On Wed, Apr 26, 2006 at 01:13:56PM -0700, Globe Trotter wrote:
> > >
> > > Has anyone used fftw on FC5? I upgraded two weeks ago, and now I get
> > > FFT's all zeros. I am not sure if I have inadvertently introduced a
> > > bug in my program somewhere, but I am mystified, where.
> > > 
> > > So, I thought I would ask if anyone has had problems/no problems
> > > with fftw on FC5.
> > > 
> > 
> > I don't know if this is of use, but since I upgraded a library that
> > uses fftw2, I've had to use -fno-strict-aliasing with gcc 4.1.0 (or
> > turn down the level of optimization below 1) in order for fftw calls
> > to work as I'd expect.  I don't know if the same is true with fftw3,
> > however.
> > 
> 
> Thanks! I think you may be right. I switched the option to
> FFTW_ESTIMATE from FFTW_MEASURE and it worked like a
> charm. Actually, it appears that FFTW_MEASURE does not work even
> with RHEL4. I have sent the FFTW developers a note asking them
> whether that needs to be called any differently.
> 

Interesting.  Would you mind posting back to this thread if/when you
hear back?  It may be worth filing a bugzilla report with the patch,
if they offer one up.

> Btw, fftw actually uses fftw3 nowadays.
> 

I did not know that.  Learn somethin' every day.

-Tim

-- 
Morals?  I eat communism and $h!t America, brother.  --Seanbaby




More information about the fedora-list mailing list