[Numpy-discussion] further questions: numpy on Solaris x86 with sun CC and libsunperf

David M. Cooke cookedm at physics.mcmaster.ca
Thu Apr 19 16:14:26 EDT 2007


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Peter C. Norton wrote:
> I have found a way to build numpy on solaris x86 with
> libsunperf. Basically, using the static library, and removing the
> compiler flag for libf77compat (I think it's deprecated and has been
> removed) and furthermore symlinking liblapack.a and libblas.a to the actual
> libsunperf.a seems to result in a successful build.
> 
> However, running numpy.test() results in the following error:
> 
> ....................................................................................F........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
> ======================================================================
> FAIL: check_large_types (numpy.core.tests.test_scalarmath.test_power)
> ----------------------------------------------------------------------
> Traceback (most recent call last):
>   File "[long path removed]/python-2.5.1c1/lib/python2.5/site-packages/numpy/core/tests/test_scalarmath.py", line 46, in check_large_types
>     assert b == 6765201, "error with %r: got %r" % (t,b)
> AssertionError: error with <type 'numpy.float96'>: got 6765201.00000000000364
> 
> ----------------------------------------------------------------------
> Ran 573 tests in 1.462s
> 
> FAILED (failures=1)
> <unittest.TextTestRunner object at 0x84a998c>
> 
> 
> Can anyone offer an opinion as to the severity of this failure?

Only serious if you're using long doubles. I'm not sure where the extra
is coming from (it's about an error of 2^-60). Could be that Solaris's
support for long doubles isn't quite good enough.

- --
|>|\/|<
/------------------------------------------------------------------\
|David M. Cooke              http://arbutus.physics.mcmaster.ca/dmc/
|cookedm at physics.mcmaster.ca
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGJ82hN9ixZKFWjRQRAhFtAJoDODg050+XppxgtfjV8iXeFEpXPQCeLuN4
fcfbssyxp0Mjry0KYnLt7p4=
=90If
-----END PGP SIGNATURE-----



More information about the NumPy-Discussion mailing list