[SciPy-user] fftpack error with large arrays

[Douglas Rudd]

Hi,

I'm trying to do a 512^3 float32 fft but I'm getting a strange error 
(output from ipython included below).  The code works up to 384^3, but 
not up to 512^3.  I've looked at the source to scipy and can't find 
where this error is being generated.

Any ideas?

Doug


In [1]: from scipy import zeros

In [2]: from scipy.fftpack import fftn

In [3]: g = zeros( (512,512,512) )

In [4]: fftn(g)
---------------------------------------------------------------------------
<type 'exceptions.ValueError'>            Traceback (most recent call last)

/Users/drudd/<ipython console> in <module>()

/sw/lib/python2.5/site-packages/scipy/fftpack/basic.py in fftn(x, shape, 
axes, overwrite_x)
     298         overwrite_x = 1
     299         work_function = fftpack.zfftnd
--> 300     return _raw_fftnd(tmp,shape,axes,1,overwrite_x,work_function)
     301
     302

/sw/lib/python2.5/site-packages/scipy/fftpack/basic.py in _raw_fftnd(x, 
s, axes, direction, overwrite_x, work_function)
     235                 x = _fix_shape(x,s[i],i)
     236     if axes is None:
--> 237         return work_function(x,s,direction,overwrite_x=overwrite_x)
     238
     239     #XXX: should we allow/check for repeated indices in axes?

<type 'exceptions.ValueError'>: dimensions too large.