Remove mode argument to GenericTensor::apply

[Martin Sandve Alnæs]

As discussed in issue#388, the mode argument to GenericTensor::apply is not actually used and should be removed.

[Martin Sandve Alnæs]

This is not true, the mode argument to GenericTensor::apply is used in petscmatrix. However it's not used in any of the other tensor classes. It's called with "add", "insert" and "flush" in various places in the code. Sounds like a potentially serious bug to me? @garth-wells: is there a reason why this is not a problem that I'm not seeing?

[Prof Garth Wells]

The flush option was added after apply was first added, and I'm not sure that flush should be performed via the apply function.

We're planning to add a Tpetra backend in the next few weeks, which may well help clarify what the behaviour of GenericTensor::apply should be.

[Martin Sandve Alnæs]

Ok. Even without flush, insert and add are fundamentally different behaviours though. Maybe just as well to make it functions 'apply_add' and 'apply_insert'.

[Prof Garth Wells]

Let's re-visit this once we have the Tpetra backend in place. Re-targeting to 1.6.

[Jan Blechta]

According to PETSc model we don't need apply_insert and apply_add but rather apply_flush and apply_final. apply_flush serves for (collective) switching between add and insert during assembly.

[Anders Logg (Chalmers)]

Removing milestone: 1.7 (automated comment)