[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [femm] Batch processing / interfacing
Dear all,
I have been reading with interest the stuff on batch processing since it is
something that would also be very useful to me, but we've got to be careful
here that we are clear what we are suggesting that Dave considers doing to
FEMM.
At the moment FEMM is extremely useful to me simply because it is so
straightforward and easy to use. I use it for teaching and for student
projects both undergraduate and postgraduate level, but I have also used it
in my research. It is ideal for these applications because it not so
overblown as some of the commercial stuff.
Without doubt a full-blown scripting structure (like FlexPDE say) would be
useful and extremely powerful but if all that is required is the facility
to run batches of solutions then there are many simpler ways.
In my old UNIX days I used to run simulation problems in batches in a
very simple way. The solver was written in such a way that it read its
input and output file names from a text file. The script which ran the
process simply told the system how many times to run the solver, where to
find the control text file and to only run the solver processes one at a
time - this so that it waited for one to finish before starting another.
It's not quite so straightforward in Windows but it may be possible to
produce a version of the FEMM solver which read from a "Batch file" the
number of problems it should solve and then the sequence of data file
names. Output file names could be specified or made up from the input name.
This means that the batch would be done in a single process not a sequence.
It would also mean that each problem has to be completely defined
beforehand and that a complete set of results would be produced for each,
but I don't see this as a problem these days. It would be nice of course to
be able to automatically step things like current density or the position
of a rotor!
Where the need for more complicated stuff occurs is in the post-processing.
Commercial stuff allows the post-processor to be run to produce a
particular result (in particular I think that the Alstom programme SLIM
allows this). For example, you could run the post-processor to do a line
integral and output the results to a file. The requirement is sent to the
processor at the command line, this is however usually a UNIX based
approach. When post processing batches in FEMM for say an animation I have
found that most time is spent in getting the picture (animation frame)
right before it's saved. If some way could be found of telling FEMMVIEW how
you wanted it to look this would be much easier.
Keith.