TODO

Immediate priorities, 3/11/2005:

1.  Break up all routines that affect equation numbering, and combine
all numbering into a single routine.  This will affect:

  Lithomop3d_setup.py
  read_bc.f
  read_wink.f
  adjid.f
  read_winkx.f

2.  Change things so that only a single element type is used at a time.
This will simplify indexing.  I need to think how this will affect
infinite elements.  This should allow us to move more quickly, but I
may need to switch back at some point, once element families are set
up better.

3.  I can probably eliminate dmat as a global matrix.  See what this
does to performance vs. memory savings.

4.  Start putting in more PETSc instrumentation (finer-grained) and
evaluating performance bottlenecks.

5.  Put in power-law viscoelastic model, and come up with equations to
implement Drucker-Prager elasto-plastic model.  Look into viscoplastic
models.

6.  Create function in array.cc to deallocate memory for arrays, and
then put in python calls for arrays that need to be deallocated.

Longer-term priorities, 3/11/2005:

1.  Create shape function info and integration routines for cohesive
and surface elements.

2.  Create routines necessary to implement cohesive (kinematic) elements
and kinematic BC elements.  This will involve routines for rotations,
use of quaternions, etc., as well as the looping over elements in a
group.

3.  Move control down a level so that routines elastc, viscos, and
iterate are done in python.


After talking to Matt and seeing if Lagrange multipliers are OK:

1.  Come up with routines for integration of surfaces/cohesive elements, analogous to solid element routines.

2.  Algorithms for setting stiffness matrix and vector structure for cohesive elements.