[gmx-developers] User Defined Input Values

[Gerrit Groenhof]

Hi Hank,

>I'm trying to modify the code slightly and am a little bit lost...
>
>I would like access to the user defined inputs (userreal1 etc.) within the
>posres() function inside of bondfree.c
>
>I think that posres() is of type ifunc (from bondf.h), and i don't think
>that the
>structure t_inputrec (which contains userreal1 etc.) gets passed to
>ifunc; but I don't want to just add it
>and mess up all the other calls to this function type.  On the other hand,
>I don't want to just clone the definition of ifunc for posre() because I
>don't really know what I'm doing and would miss alot of inclusions and
>stuff
>
>what is the cleanest solution for gaining access to the user inputs inside
>the posres function?  Or are they already there and I haven't recognized
>them?

>Thanks for any help and sorry if my message isn't clear, programming isn't
>my specialty :P

I'm not completely sure wheter this is what you need, but maybey it
is:

You only want to use the userint, and userreals, right?

In that case the easiest, but maybe not the cleanest, way would be to
add userreal and userint elements to one of the structs that is passed
on to posres, like the forcerec for instance. Then you can fill this
userelement(s) when the forcere struct itself gets initiated and
filled by init_forcerec (force.c) with the parameters stored in the
inputrec. The user variables get read into the inputrec by grompp, but
are not used furtheron in the code. The good thing is that if a user
wants to use some extra input variables they he uses the userint/reals
and doesn't have to modify the horrible(!!) grompp code.

If you also want to get acces to usergroups, it is slightly different
and more complicated.

Acces to any group defined in your mdp file (Tempcoupl. grp,
energygrp, accel. grp, freeze grp, user1 grp, user2 grp, com motion
grp, XTC grp and orientation orientation grp) you get through the
t_mdatoms struct.

the mdatoms struct (include/types/mdatom.h) contains a array for each
group (cTC[], cENER[], cACC[], cFREEZE[], cUser1[], cUsre1[], cVCM[],
cXTC[], cORFIT[], the order here reflects the order above). These
arrays, the size of each is equal to the total number of atoms, are
just handles to determine to which group a certain atoms belongs. I
will explain that with an example below.

The number of different groups of each group type (Tempcoupl. grp,
energygrp, accel. grp, freeze grp, user1 grp, user2 grp, com motion
grp, XTC grp and orientation orientation grp) is stored the atoms
struct
(include/types/atoms.h) of the topology struct
(include/types/topology.h)

It is a bit difficult to explain, so I use an example:

Suppose you have defined two freeze groups groups in your mdp file:
atomsA and atomsB.

Then the topology.atoms.grps[egcFREEZE] is 3. This means there are
3(!) freeze groups, namely, atomsA (0), atomsB (1) and the rest
(3). This is also printed to the output by grompp by the way.

The elements of the mdatoms.cFREEZE[nratoms] array in mdatoms struct
(include/types/mdatom.h) are either zero, one or two. If an atom i has
md.cFREEZE[i]==0, it belongs to the first freeze group, atomsA. if it
has md.cFREEZE[i]==2, it does belong to the rest (and by definition it
is not a freeze group, so its positions will be updated in the code).

I think if you follow this, you are there.  So to get acces to atomsA,
the first freeze group, you have to loop over all atoms and scan for
the ones that have md.cFREEZE[i]==0 and store them in some other array
or do whatever you want with them. To get acces to atomsB, you have to
check for md.cFREEZE[i]==1, and the rest you get by scanning for
md.cFREEZE[i]==2.

The mdatoms struct is passed on to alomst all important functions, as
it contains the information mdrun needs to do the dynamics. Therefore
you can do the scan trick almost everywhere.

I hope I helped you a bit in the right direction,

good luck,

Gerrit