[SciPy-user] installation on Suse8.1 (blas problem)

Pearu Peterson pearu at cens.ioc.ee
Mon Mar 17 07:42:16 EST 2003


On Mon, 17 Mar 2003 baecker at physik.tu-dresden.de wrote:

> I followed the approach to append the missing symbols to
> skip_names.
> 
> So I did the following in /Lib/linalg/setup_linalg.py:
> a) skip_single_routines = 1
> b) further down I finally ended up with
>            skip_names['fblas'].extend(\
>             'drotmg drotm  srotg crotg srotmg srot csrot srotm sswap cswap
> sscal cscal'\
>             ' csscal scopy ccopy sdot cdotu cdotc snrm2 scnrm2 sasum
> scasum'\
>             '  sgemv cgemv chemv ssymv strmv ctrmv'\
>             ' sgemm cgemm'.split())

The same effect (in b) can be now achived by setting
  using_lapack_blas=1
in the header of Lib/linalg/setup_linalg.py. 
Let me know if it works for you.

> However, for a scipy.test(10) I get the following error message:
> 
> ======================================================================
> ERROR: bench_random (test_basic.test_irfft)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File
> "/home/baecker/PYTHON/lib/python2.2/site-packages/scipy/fftpack/tests/test_basic.py",
> line 410, in bench_random
>     print '|%8.2f' % self.measure('FFT_irfft(x1,size)',repeat),

<snip>

> "/home/baecker/PYTHON/lib/python2.2/site-packages/Numeric/FFT/FFT.py",
> line 36, in _raw_fft
>     wsave = init_function(n)
> TypeError: an integer is required

This could be a Numeric issue. What Numeric version are you using?
(I have tested fftpack benchmark tests to pass upto Numeric 22.0, I
haven't tried 23.0, yet)

> There are some missing fields concerning fft, so should I expect
> the above error ?

No.

> dfftw_info:
>   FOUND:
>     libraries = ['drfftw', 'dfftw']
>     library_dirs = ['/usr/lib']
>     define_macros = [('SCIPY_DFFTW_H', None)]
>     include_dirs = ['/usr/include']

This looks good to me. Since scipy.test(1) includes fftpack sanity tests
that all pass, then scipy.fftpack should be ok.

Thanks for your feedback,

Pearu




More information about the SciPy-User mailing list