[gmx-users] Integrating with BOINC: failing at minimization

[Mark Abraham]

Jack Shultz wrote:
> Hello,
> 
> I've tried a couple ways of running the workshop tutorial with a
> different protein and amber03 force field. It works with a clean
> install of gromacs, but I am trying to integrate it into a grid
> computing project.
>  https://extras.csc.fi/chem/courses/gmx2007/tutorial1/tutorial1.pdf
> 
> I tried several approaches to running this app but encountered a
> strange problem in both approaches to integrating. It always has
> trouble minimizing. This indicates to me, there is something strange
> with the environment that I created. If you have any hints for me that
> would be great. The two approaches I tried were:
> 1) putting all the dependent files in the same directory as the binary
> and executing it, this failed when I tested it stand alone and it
> created this same error
> 2) I tried writing a bash script to mimics the GMXRC.bash except it
> maps every directory relative to a temporary directory where
> everything runs. (e.g. GMXBIN=../slot/0/bin/)
> It seems to run fine through a workunit but then it does not minimize
> properly causing additional problems downstream
> 
> 
> Steepest Descents converged to machine precision in 54 steps,
> but did not reach the requested Fmax < 10.
> Potential Energy  = -8.35030867755191e+214
> Maximum force     = 7.83437759695905e+221 on atom 25439
> Norm of force     = 5.13628303396235e+219
> 
> http://www.hydrogenathome.org/result.php?resultid=1341858
>

grompp is giving a bunch of warnings about long bonds. If you really are 
doing the same calculation you did with a "clean install of GROMACS" 
then they were also given there. You should attend to them, since your 
output is consistent with http://wiki.gromacs.org/index.php/blowing_up. 
You need to fix that before you can isolate whether you have other problems.

Mark