[gpaw-users] Test failures

[Ondrej Marsalek]

2010/11/16 Jens Jørgen Mortensen <jensj at fysik.dtu.dk>:
>
> 10.10 also works for me out of the box with atlas-sse2.  Have you tried
> to compile GPAW with atlas instead of MKL?  If yes, what atlas library
> did you use - maybe try the generic one.

Indeed, on Ubuntu 10.10, with an empty customize.py, openmpi 1.5, gcc
and atlas from the repositories, I get a build that passes all the
test. Intel is really starting to bug me. It was my impression that
problematic builds with Intel tools were only specific to another
project that I shall not name here.

A bit strangely, Two tests failed on first pass and passes when i
reran the testsuite with -f. The original errors are cited below is it
is of interest to anyone. I understand that stark_shift is just a
numerical difference, but don't really know what the problem was in
lcao_projections.

Best,
Ondrej


parallel/lcao_projections.py    12.346  FAILED!
#############################################################################
Traceback (most recent call last):
  File "/home/andy/build/gpaw/gpaw/test/__init__.py", line 350, in run_one
    execfile(filename, loc)
  File "/home/andy/build/gpaw/gpaw/test/parallel/lcao_projections.py",
line 24, in <module>
    assert np.allclose(eig, eig_ref)
AssertionError
#############################################################################

stark_shift.py                  69.163  FAILED!
#############################################################################
Traceback (most recent call last):
  File "/home/andy/build/gpaw/gpaw/test/__init__.py", line 350, in run_one
    execfile(filename, loc)
  File "/home/andy/build/gpaw/gpaw/test/stark_shift.py", line 213, in <module>
    assert abs(pol1+pol2) < 0.0001
AssertionError
#############################################################################