[gmx-developers] innerc to forcebuf: Energy is still zero

[Erik Lindahl]

Hi,

>
>
>Problem is, I don't get the pair-potential energies (they are zero all
>the time) in output files for bham and new interaction. Strange thing is
>I didn't even touch bham loops. Moreover,
>print statements at inner loop for vnbtot gives no-zero values at all. 
>

vnbtot and vctot are only the temporary variables used in the innermost 
loop; at the
end of the outer loop the are assigned to Vc(ggid) and Vnb(ggid), where 
ggid is the group
index. If you only have one energy group ggid=0.

Check that these two variables are non-zero, and if that's the case you 
can also check
in fnbf.c (after calling the inner loop) that the corresponding arrays 
egnb(ggid)
and egcoul(ggid) have the energy.

>
>I have checked in may places that energy arrays are zero.(a 14,15th
>members where defined as bham and new pot's energies)
>- before and after or inside: do_fnbf(), do_force(), sum_epot() 
>  grps->estat.ee[egBHAM] and   ener[i]
>- identical input files with original GROMACS code gives energies
>properly as expected :-)
>
>Apparently, either energies do not passed properly from inner loops or I
>made
>some error in modification that affects this mechanism.
>
>How could I put a print statements in outer-loops? Cos, what I did for
>innerloops is
>defining new function(simple print statement) and use this function with
>assign() in
>mkinl routines.In outher loops this strategy didn't work cos compiler
>complains about conflicting data types.
>

What I usually do when I debug is that I first create the inner loop in 
C or Fortran
by going to gmxlib and issuing e.g 'make innerc.c'. Then I can edit the 
loops I
want like a normal routine, and when you issue 'make' again the modified 
loop will
be newer than the source files used to create it, so it will not be 
overwritten.

Cheers,

Erik