spott / erwin Goto Github PK

Mat is really mat_t** or something similar. We should cary around a mat_t* instead, to reduce overhead.

The left and right eigenvectors of the non-hermitian hamiltonian need to be mutually orthogonalized.

So we want L^H R = I.

Or write our own assert to have a message as well.

More descriptive error messages would be really nice.

This should speed things up significantly.

The non uniform grid needs to be written. Specifically a 1/r -> linear grid.

higher order derivative support is probably necessary for grids without first/second derivative matching.

Support the symmetric/hermitian matrix type from PETSc for hermitian hamiltonians.

Static variables are global, not unique to scope. So they should be removed from everywhere.

Including lambdas. A better method is "mutable" lambdas that modify captured variables. This does more of what is thought.

Different parameter sections are running into problems:

Different parameter sections for propagation, which are supposed to be independent, require the same information (folder is one, maybe others?). What is the best way to handle this? (should write() instead take a parameter folder?).

LaserParameters::sin_squared needs the frequency to figure out the period. How should this class get this information?

One option is to make envelope take a frequency...