[SciPy-user] Absoft Compiler

[Travis Oliphant]

> Well, a mere four months after the SciPy conference in Pasadena (thanks
> all!), I have finally built SciPy.  I include the output of some of the
> diagnostic commands after my signature, but to summarize:  I built
> LAPACK, ATLAS, and SciPy from source using the Absoft F77 compiler v4.6
> and gcc v3.2 on RedHat 8.0.  scipy.test(level=1) fails with the
> following error:
>
> Traceback (most recent call last):
>   File
> "/usr/lib/python2.2/site-packages/scipy/stats/tests/test_morestats.py",
> line 30, in check_basic
>     assert_almost_equal(w,0.90047299861907959,7)
>   File "/usr/lib/python2.2/site-packages/scipy_test/testing.py", line
> 335, in assert_almost_equal
>     assert round(abs(desired - actual),decimal) == 0, msg
> AssertionError:
> Items are not equal:
> DESIRED: 0.90047299861907959
> ACTUAL: 0.90047270059585571
>

Congradulations.  This error should be ignored.  Basically, we are
requiring too many matching digits (the routine uses single-precision and
we should only be interested in 6 significant digits).  This is changed.


> There are a couple of additional problems:  the SciPy install didn't
> detect that I have the fftw libraries installed (from the RPM), and the
> test suite complains up near the top about the lack of a test_matfuncs
> module.

There is a site.cfg file where this information can be placed if it is in
an unusual location.

Welcome to the group....


-Travis O.