~pefarrell/dolfin-adjoint/alnaes-replay-cb : changes

~pefarrell/dolfin-adjoint/alnaes-replay-cb » Changes from revision 641

From Revision 641 to 622

Rev	Summary	Authors	Date
641	Implement a replay_cb in the ReducedFunctional to let users ... Implement a replay_cb in the ReducedFunctional to let users look at convergence of the state.	Patrick Farrell	11 years ago
640	Handle Constant.assign. Handle Constant.assign. There are two options how to handle this. (Well, three, really). The first option is to just consider Constants like Functions in FunctionSpace(mesh, "R", 0), and then all the machinery just works: they are annotated as initial conditions, their updates get assignment equations, etc. The disadvantage of that is the unacceptable performance penalty: most likely Constants will be used in many equations, and so their adjoint equation will have lots and lots of entries, and be a nightmare to assemble (c.f. Marie's experiences of the integral-constraint Lagrange multiplier in the cardiac example). It isn't necessary at all to compute it, but it would be computed if they were treated like Functions. The second option (adopted) is to consider the value of Constants as a mutable part of the state to be bundled with the context of the callbacks registered with libadjoint, just as the values of parameters to Expressions are. This slightly increases the memory required, but hopefully people aren't declaring millions of Constants. It has a nasty side-effect which I'll discuss below. The third option (preferred) is to not assign to Constants. They're either Constant or they're NotConstant! A Constant that changes value is a contradiction in my book, but anyway. I don't get to define dolfin semantics. The nasty side effect: so, when a callback is called, it re-sets the values of all Constants to the values they had when it was registered -- as described above, the values of Constants are now part of the state to be re-set. Except when they're not! And when are they not? When you're perturbing the value of the damn Constant (such as for a Taylor test, or an optimisation). Then you don't want to re-set the Constant value! Agh! You only want to re-set the values of the "other" constants. Hideous, no? Anyway, the best solution I could come up with was to say that "Constants that are part of a ScalarParameter aren't to be re-set", which fixes the Taylor test/optimisation problem at the cost of a nauseating hack. An alternative would be to introduce a NotConstant class, which behaves like a Constant except it can be assigned to, but I doubt that would fly.	Patrick Farrell	11 years ago
639	Some fixes for symmetric BC application and dolfin == 1.2.0. Some fixes for symmetric BC application and dolfin == 1.2.0.	Patrick Farrell	11 years ago
638	Expose some utility functions useful for annotating nonlinea... Expose some utility functions useful for annotating nonlinear problems with Picard: annotate, and record.	Patrick Farrell	11 years ago
637	Not called symmetric_operator anymore. Not called symmetric_operator anymore.	Patrick Farrell	11 years ago
636	Set the parameters in the replay too. Set the parameters in the replay too.	Patrick Farrell	11 years ago
635	Get the replay for the PointIntegralSolver working. Get the replay for the PointIntegralSolver working.	Patrick Farrell	11 years ago
634	Annotate PointIntegralSolver.step(). Replay not working yet. Annotate PointIntegralSolver.step(). Replay not working yet.	Patrick Farrell	11 years ago
633	Add a small ODE solver test case to get working Add a small ODE solver test case to get working	Patrick Farrell	11 years ago
632	Fix bug lp:1171322, reported by Marie Rognes. Fix bug lp:1171322, reported by Marie Rognes.	Patrick Farrell	11 years ago
631	Fix bug lp:1170334 reported by Gabriel Balaban. Fix bug lp:1170334 reported by Gabriel Balaban.	Patrick Farrell	11 years ago
630	Some updates to the wrap_solve. Some updates to the wrap_solve.	Patrick Farrell	11 years ago
629	Allow dolfin-adjoint to use solver tolerances set by the use... Allow dolfin-adjoint to use solver tolerances set by the user, even though the lack of an identity operator forces us to use solve(A, x, b).	Patrick Farrell	11 years ago
628	Zero pad the output of hessian_eps. Zero pad the output of hessian_eps.	Patrick Farrell	11 years ago
627	Annotate the Hessian action with timers Annotate the Hessian action with timers	Patrick Farrell	11 years ago
626	Happily abort optimisation tests if no suitable scipy versio... Happily abort optimisation tests if no suitable scipy version is installed	Simon Funke	11 years ago
625	Work around UFL bug in 1.2.0. Work around UFL bug in 1.2.0.	Patrick Farrell	11 years ago
624	Fail gracefully if SLEPc is unavailable. Fail gracefully if SLEPc is unavailable.	Patrick Farrell	11 years ago
623	Specify the perturbation direction to stop minor test tolera... Specify the perturbation direction to stop minor test tolerances tripping us up.	Patrick Farrell	11 years ago
622	Set MUMPS in the Riesz-representation solve also. Set MUMPS in the Riesz-representation solve also.	Patrick Farrell	11 years ago

Older »