Also report time per timestep, and percentages.

Also use this to make a checkpoint write / restore / compare test.

This makes parameters harder to find in the documentation than is necessary.

Clone the lattice to preserve previous time steps? This may be useful.

• LB init (2d/3d)
• Mask init (2d/3d)
• Elec init (2d/3d)
• Also generate example pictures from this.

Should be able to render directly in Paraview somehow... Can maybe use secondary xdmf files?

Check for NaNs... anything else?

Running Doxygen with the current Doxyfile produces empty output.

• Split it up or reorder?
• More comments
• Nicer logging

Can maybe use @eXchange, but probably have to add a new one - or the same as the one that identified globals that need to be included in the checkpoint.

Parameters need to be spread out over different structs to make input file more readable.
New MPI types need to be built.
MPI types for default structs have to be added and broadcast.

• Have all fields defined in lattice.d.
• Have all calculation routines in their respective modules:
  • density -> density.d
  • psi -> force.d
  • bdist2 -> collision.d or eqdist.d

Test passes on local machine, fails with Travis.

Check for structure of gcc matrix:

1. Upper triangular: only run half the combinations, multiply psifun1 by 0.5 when not on the diagonal.
2. No non-zero entries in diagonal, do not run the diagonal (probably not worth it, the continue statement is fast enough).

Introduce documented wrapper functions that call the mixin functions.

Currently the equilibrium distribution does not conserve mass.

This needs to be added. Loading it as a dynamic library causes complications...