[Numpy-discussion] svn test failures: linalg with complex

[Pauli Virtanen]

Sat, 18 Apr 2009 09:41:28 -1000, Eric Firing wrote:

> I updated my Thinkpad to Ubuntu 9.04 release candidate, 32-bit, got a
> fresh svn checkout of numpy trunk, built, and installed.  Running
> numpy.test() yields 16 errors (output attached) in linear algebra on
> complex arguments.  Is this expected?

The Numpy test suite should run without any failures. The failures here 
likely indicate that something is wrong with LAPACK/BLAS/ATLAS on
Ubuntu 9.04.

Apparently, at least ATLAS's complex single-precision routines are 
completely hosed on that platforms. I filed a bug here:

	https://bugs.launchpad.net/ubuntu/+source/atlas/+bug/363510

Seems specific to the SSE version of Atlas. Could you comment on the bug 
tracker if you see the same thing on your machine?

Consequently, on 9.04 I get the following failures, but for some reason 
these are not the same as yours:


Running unit tests for 
numpy-1.4.0.dev6826.4e95223c11c627aa8f4878b40953d75ea8228add-py2.5-linux-
i686.egg.numpy
NumPy version 1.4.0.dev6826.4e95223c11c627aa8f4878b40953d75ea8228add
NumPy is installed in /home/pauli/koodi/proj/scipy/numpy.git/dist/linux/
lib/python2.5/site-packages/
numpy-1.4.0.dev6826.4e95223c11c627aa8f4878b40953d75ea8228add-py2.5-linux-
i686.egg/numpy
Python version 2.5.4 (r254:67916, Apr  4 2009, 17:55:16) [GCC 4.3.3]
nose version 0.10.4
.......
======================================================================
FAIL: test_csingle (test_linalg.TestLstsq)
----------------------------------------------------------------------
Traceback (most recent call last):
  File ".../numpy/linalg/tests/test_linalg.py", line 39, in test_csingle
    self.do(a, b) 
  File ".../numpy/linalg/tests/test_linalg.py", line 141, in do
    assert_almost_equal(b, dot(a, x))
  File ".../numpy/linalg/tests/test_linalg.py", line 23, in 
assert_almost_equal
    old_assert_almost_equal(a, b, decimal=decimal, **kw)
  File ".../numpy/testing/utils.py", line 262, in assert_almost_equal
    return assert_array_almost_equal(actual, desired, decimal, err_msg)
  File ".../numpy/testing/utils.py", line 537, in 
assert_array_almost_equal
    header='Arrays are not almost equal')
  File ".../numpy/testing/utils.py", line 395, in assert_array_compare
    raise AssertionError(msg)
AssertionError:
Arrays are not almost equal

(mismatch 100.0%)
 x: array([ 2.+1.j,  1.+2.j], dtype=complex64)
 y: array([  7.22953510+1.74995685j,  10.88255692+1.7498703j ], 
dtype=complex64)

======================================================================
FAIL: test_csingle (test_linalg.TestSolve)
----------------------------------------------------------------------
Traceback (most recent call last):
  File ".../numpy/linalg/tests/test_linalg.py", line 39, in test_csingle
    self.do(a, b)
  File ".../numpy/linalg/tests/test_linalg.py", line 76, in do
    assert_almost_equal(b, dot(a, x))
  File ".../numpy/linalg/tests/test_linalg.py", line 23, in 
assert_almost_equal
    old_assert_almost_equal(a, b, decimal=decimal, **kw)
  File ".../numpy/testing/utils.py", line 262, in assert_almost_equal
    return assert_array_almost_equal(actual, desired, decimal, err_msg)
  File ".../numpy/testing/utils.py", line 537, in 
assert_array_almost_equal
    header='Arrays are not almost equal')
  File ".../numpy/testing/utils.py", line 395, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
Arrays are not almost equal

(mismatch 100.0%)
 x: array([ 2.+1.j,  1.+2.j], dtype=complex64)
 y: array([  7.22953510+1.74995685j,  10.88255692+1.7498703j ], 
dtype=complex64)

----------------------------------------------------------------------
Ran 2034 tests in 41.154s

FAILED (KNOWNFAIL=1, failures=2)

-- 
Pauli Virtanen