Implement custom integrals in uflacs

Comments (11)

Martin Sandve Alnæs reporter
Issue ~~#103~~ was marked as a duplicate of this issue.
- 2016-09-26T13:15:43+00:00
Martin Sandve Alnæs reporter
- assigned issue to
  
  Martin Sandve Alnæs
- 2016-09-30T14:48:43+00:00
Martin Sandve Alnæs reporter
Some progress now, class names are easier to get after my latest work on jit dependencies.

@logg what do you think of this idea:

Compute tables with some quadrature degree N>0, use those to to determine nonzero dof ranges like the other code generation does. This will make a lot of existing code paths unchanged, but it relies on no single nonzero basis function being zero in all quadrature points. That sounds safe to me, even for N=1, but maybe pick N=3 to be on the safe side.
- 2016-09-30T14:56:13+00:00
Martin Sandve Alnæs reporter
- changed milestone to 2017.1
- 2016-10-13T06:45:10+00:00
Jan Blechta
- changed milestone to 2017.2
- 2017-04-21T19:24:59+00:00
Martin Sandve Alnæs reporter
- removed responsible
- 2017-06-19T14:22:32+00:00
Jan Blechta
- assigned issue to
  
  Anders Logg (Chalmers)
- changed milestone to 2018.1
- marked as blocker
This issues blocks removing the quadrature representation. Preliminary assigning to Anders.
- 2017-12-07T08:51:59+00:00
Anders Logg (Chalmers)
This feature has been implemented already by @magneano and is waiting to be merged. Also relies on this PR for UFL: https://bitbucket.org/fenics-project/ufl/pull-requests/83/avoid-lowering-facetnormal-to/diff
- 2017-12-07T11:39:28+00:00
Jan Blechta
Where is it waiting?
- 2017-12-11T18:26:31+00:00
Anders Logg (Chalmers)
It's waiting for this to be merged into UFL https://bitbucket.org/fenics-project/ufl/pull-requests/83/avoid-lowering-facetnormal-to/diff
- 2017-12-11T18:51:11+00:00
Miklós Homolya
It is possible to file the FFC merge request before all the dependencies are merged.
- 2017-12-12T16:57:41+00:00
Log in to comment