ffc team mailing list archive

[ffc@xxxxxxxxxx (FFC)]

One or more new changesets pushed to the primary ffc repository.
A short summary of the last three changesets is included below.

changeset:   1484:6ff16575e6303395ce5ea524a65d0234cecf7f88
tag:         tip
user:        Johannes Ring <johannr@xxxxxxxxx>
date:        Fri Mar 27 09:10:08 2009 +0100
files:       setup.py
description:
Fix in setup.py for Windows.


changeset:   1483:de77c75d92150de0dae9480dc189bd7ad493f6cb
user:        Anders Logg <logg@xxxxxxxxx>
date:        Thu Mar 26 17:15:55 2009 +0100
files:       test/simple_verify_tensors/test.py
description:
Add comment about which tests should fail. Everything looks OK now!!!


changeset:   1482:fd13957f73b480d4b86c605b24e2d5265aec91e8
user:        Anders Logg <logg@xxxxxxxxx>
date:        Thu Mar 26 14:40:35 2009 +0100
files:       test/simple_verify_tensors/reference.pickle test/simple_verify_tensors/test.py
description:
Bug fix in simple_verify_tensors and update reference values.
All looks ok now. Three demos should fail with quadrature.
Comments from Kristian:

FunctionOperatorsBilinearForm_cell_integral_0:    *** (diff = 0.010091)

  The reason for the 'error' is that the old FFC interprets

  1/f := basis*1/w[*][*]

  where the new FFC (quadrature) does:

  1/f = 1/(basis*w[*][*])

MassBilinearForm_cell_integral_0:                 *** (diff = 1e-09)

  The code is correct, the old quadrature has a similar discrepancy. I think it is
  because we have 64 quadrature points (tabulated values) that are multiplied and
  the tensor contraction does this in Python and only used one value for the
  basis in the generated code.

QuadratureElementBilinearForm_cell_integral_0:    *** (diff = 0.249892)
QuadratureElementLinearForm_cell_integral_0:      *** (diff = 0.10106)

  This is because the old FFC interprets element = QE("triangle", 3) as a 3x3
  quadrature scheme, but for the new FFC '3' is the order and it will use a 2x2
  scheme for the evaluation.

----------------------------------------------------------------------
For more details, visit http://www.fenics.org/hg/ffc