On Wed, Jun 16, 2010 at 11:36:16AM +0200, be...@ch... wrote:
> Hello Ulf
Hi,
I just assume you do not mind if I make this public (if you do, write me
an angry email and I will stop).
> Thank you for your advises. You asked for wishes for the new version of
> wavepacket (for a higher acceptance in the research).
> One wish would be the implementation of optimal control. Some years ago I
> already implemented a simple Tannor scheme for forward and backward
> propagation. Since now backward propagation is directly possible, it
> should be possible as well to implement Optimal Control in the wavepacket
> program to compete with the MCTDH package, were this is already
> implemented.
Good point. The idea has been around for quite some time, and I think we
(at least I) never implemented it because it is a bit cumbersome to do
in the stable branch. Essentially all the ideas we had included hacking
qm_propa in a most ugly way.
However, if I remember correctly, the only really missing functionality
is the ability to do some custom calculations during the propagation.
This can be implemented in principle (add a function reference that gets
called at each main time step), and the other obstacles can be overcome
with some scripting...
In short: Added to the TODO. It will result in a not too elegant hack
(something like the scripts for explicit pump-probe calculations, see
http://wavepacket.sourceforge.net/Demos/FemtoChem/Interferometry), but
it should be possible to write a script that can be applied to general
problems.
> Another point is the possibility of creating movies in a non-interactive
> way, i.e. to let run qmpropa in a queue, where a movie is created in a
> file, which can be looked at after the job finished. I think today, one
> always has to run the program interactively and to shut down the
> screensaver to make a movie of the wavepacket propagation.
The problem here is that the way Matlab creates a movie is letting you
paint the image, then making a screenshot. I can have a look around if
there are packages that have a different approach, but I would not count
on it.
However, the qm_movie script might be able to circumvent this problem.
With this, you would do the calculations on the single nodes
non-interactively (turn plotting off), save the results, and then use
the qm_movie script afterwards to extract the films from the
calculations (which requires interactive access, but goes relatively
fast).
This still poses problems if your calculations are too large to be
saved, though. Is this the case for you?
Regards,
Ulf
|
