[gmx-users] "B" state for free energy corrupts minimization?

[mernst at tricity.wsu.edu]

>Indeed the zeros in the bonds (and probably also in the angles?) are the
>problem.
>This is caused by a bug that I have fixed in November,
>but we did not release 3.3.1 yet:
>http://www.gromacs.org/pipermail/gmx-revision/2005-November/000117.html
>
>Apparently some bonded interaction parameters are not defined
>for bonded interactions that involve your 36 atom type in the B-state.
>
>You should add parameters for these bonded interactions,
>either directly after the definition of the bonded interaction,
>or in the force-field bondtypes and angletypes sections.
>
>Berk.

Berk,

Thank you for the hint. I was able to fix the zeros in the bonds, but I am having
trouble tracking down the source of the zeros in the angles. Which file or files needed
modification to fix this bug? I have been using 3.3. I tried downloading the source and
building from CVS, but it failed:

domdec.c: In function `dd_move_x':
domdec.c:143: `mpi_req' undeclared (first use in this function)
domdec.c:143: (Each undeclared identifier is reported only once
domdec.c:143: for each function it appears in.)
domdec.c:143: too many arguments to function `low_dd_move_x'
domdec.c: In function `dd_start_move_x':
domdec.c:155: structure has no member named `mpi_req'
domdec.c:155: too many arguments to function `low_dd_m

If I could patch 3.3 or build just the tool I need from CVS (grompp?) then I think I
would be able to track down the parameters I need to supply more easily.

Matt Ernst
Washington State University