~simon-funke/dolfin-adjoint/dolfin-adjoint-dirichlet-bc : changes

~simon-funke/dolfin-adjoint/dolfin-adjoint-dirichlet-bc » Changes from revision 857

From Revision 405 to 386

Rev	Summary	Authors	Date
405	Update the license .. needs to be LGPL v3. Update the license .. needs to be LGPL v3.	Patrick Farrell	12 years ago
404	A minor tolerance tweak A minor tolerance tweak	Patrick Farrell	12 years ago
403	Check the ratio is correct, even when using a mass matrix Check the ratio is correct, even when using a mass matrix	Patrick Farrell	12 years ago
402	Make the gst_mass test run in parallel. Make the gst_mass test run in parallel.	Patrick Farrell	12 years ago
401	Turns out we shouldn't use numpy's inner product in parallel... Turns out we shouldn't use numpy's inner product in parallel.	Patrick Farrell	12 years ago
400	Always patch ufl's Form.__mul__ and __rmul__. The test doesn... Always patch ufl's Form.__mul__ and __rmul__. The test doesn't work in parallel for large numbers of processors (the mesh is too small, some processors get no mesh, and then it crashes) Hopefully this hack can be removed at a later date, once a version of UFL is widespread with the necessary change.	Patrick Farrell	12 years ago
399	Merge the work in the dolfin-adjoint gst branch. The general... Merge the work in the dolfin-adjoint gst branch. The generalised stability theory calculations can now use norms other than the vector l2 norm. The GST defaults to using the mass matrices for the relevant spaces (as that's the obvious finite element thing to do), but users can pass other bilinear forms to measure size if they wish.	Patrick Farrell	12 years ago
398	The buildbot doesn't seem to like UnitInterval for some reas... The buildbot doesn't seem to like UnitInterval for some reason -- let's try UnitSquare.	Patrick Farrell	12 years ago
397	Merge the time-functionals branch. Merge the time-functionals branch.	Patrick Farrell	12 years ago
396	Merge again from dolfin-adjoint-optimisation branch Merge again from dolfin-adjoint-optimisation branch	Simon Funke	12 years ago
395	Merge optimisation branch Merge optimisation branch	Simon Funke	12 years ago
394	Warn the user if he/she creates two functions with the same ... Warn the user if he/she creates two functions with the same name	Simon Funke	12 years ago
393	Actually use the recorded initial value for Newton solves Actually use the recorded initial value for Newton solves	Simon Funke	12 years ago
392	Add new routine "find_previous_variable" to find the previou... Add new routine "find_previous_variable" to find the previous version of libadjoint variable (in the same or earlier timesteps). This routine is already used in order to find the correct depedency variable for the Newton initial condition	Simon Funke	12 years ago
391	Add a new debugging flag the checkpointing: replay = True co... Add a new debugging flag the checkpointing: replay = True compares all the revolve replays with the initial forward model solutions	Simon Funke	12 years ago
390	Bugfix in the fussy replay for nonlinear solves: make sure t... Bugfix in the fussy replay for nonlinear solves: make sure that the annotated initial condition always refers to the latest variable iteration	Simon Funke	12 years ago
389	Fix bug in checkpointing logic: honor the checkpoint flag wh... Fix bug in checkpointing logic: honor the checkpoint flag when registering initial conditions. An additional check is required to make sure that only variables are checkpointed that actually exist in libadjoints annotation.	Simon Funke	12 years ago
388	Add a test where the user sneakily accesses a .vector() by h... Add a test where the user sneakily accesses a .vector() by hand and changes it. I think we should try to catch this case: dolfin-adjoint should either work or give a useful error message, not fail silently.	Patrick Farrell	12 years ago
387	Now that we're not bumping the timestep automatically anymor... Now that we're not bumping the timestep automatically anymore, change the warning so that it's actually true.	Patrick Farrell	12 years ago
386	If we try u.assign(u), it will screw up the annotation. In s... If we try u.assign(u), it will screw up the annotation. In such a case, just do nothing.	Patrick Farrell	12 years ago