lp:~fluidity-core/fluidity/fluidity_libadjoint_navier_stokes

Branch merges

Propose for merging

No branches dependent on this one.

Merged into lp:~fluidity-core/fluidity/adjoint at revision 3603

Simon Funke: Needs Resubmitting on 2011-11-11

Patrick Farrell: Pending requested 2011-11-07

Diff: 215907 lines api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

3774. By Simon Funke on 2012-03-26

Merge from trunk

3773. By Patrick Farrell on 2012-03-24

Merge the recent change to ADJ_VEC_GET_SIZE_CB.

3772. By Simon Funke on 2011-12-22

Some minor changes. All short tests pass and shallow_water_mms_wave

3771. By Simon Funke on 2011-12-22

Extend the cube-1-replay test so that it tests that the replay produce bitwise the same answer as the original run

3770. By Patrick Farrell on 2011-12-22

Make the cube-1-replay test check the end-time of the forward replay.

3769. By Patrick Farrell on 2011-12-22

Add the .msh file to the cube-1-replay test so that it runs through testharness.

3768. By Simon Funke on 2011-12-22

Correct the owner of the revolve tests and add the adjoint flag to the cube 1 test

3767. By Simon Funke on 2011-12-22

A initial version of a automatic annotation test

3766. By Simon Funke on 2011-12-22

Finally a bitwise correct replay of the cube-1 example.

Before, the timefield was not correct because it is calculated as a diagnostic in the forward model.
For simplicity, I had the automatic annotation disabled for diagnostic caluclation. So all I had to do
is to switch on the automatic annotation, which then gave immediatly the correct answer.

The cube-1 example is sligtly modified:
- It uses LU solver instead of iterative solvers. This is because preconditioners are not passed into the compute_forward loop (yet).
- The reference node in the pressure solver option is not used (simple because I the annotation is not coded up yet)
- The pressure guess option is set to "never" as opposed to "only first timestep". Again, the reason is to avoid coding up the required annotation.

3765. By Simon Funke on 2011-12-22

After replacing the dummy callback for set_scalar_boundary_conditions_values with a correct implementation, we now get bitwise correct pressure and velocity values for both timesteps.