[gmx-users] Re: Re; Failed Energy Minimization (solved!)

[Erik Lindahl]

Hi Justin,

Double precision energy minimization works great for us, so if you  
have system where you think there _might_ be a bug, please submit it  
to bugzilla.

In fact, you should always use double precision l-bfgs mimimization  
prior to normal mode analysis, IMHO...

Cheers,

Erik


On Apr 20, 2007, at 9:02 PM, Justin M. Shorb wrote:

> Greetings everyone!
>
> After trying my same shell scripts on another computer, using  
> single precision, I have worked out that the double precision  
> version of Gromacs-3.3.1 does not energy minimize consistently. The  
> exact same run script using different force fields will energy  
> minimize sometimes, but most force fields give infinite forces.  
> This was verified for diglycine, trialanine, alanine dipeptide as  
> well as a very large protein. In fact, the exact same run script  
> for diglycine gives minimization within 8 steps for single  
> precision, where with double precision it had infinite forces (for  
> no reason) in step 0. I even wrote a quick script to determine the  
> nearest neighbors for each atom, and found that all atoms were at  
> least 0.9 Angstrom apart (so no overlap).
>
> Recompiling the exact same source code in single precision on my  
> own machine reproduced the success identically. Thus I am left to  
> believe that there is bug in the double precision version of the  
> energy minimization (mdrun_d or grompp_d, I don't know) code.
>
> Just thought I would let everyone know to avoid double precision  
> when running energy minimization.
>
> Thanks for all your suggestions!
> Justin
>
> _________________________________________________________________
> Justin M. Shorb			Phone: (608) 262-0483
> Skinner Group				shorb at wisc.edu
> University of Wisconsin-Madison,  Department of Chemistry
> 1101 University Ave., Madison, WI, 53706
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before  
> posting!
> Please don't post (un)subscribe requests to the list. Use the www  
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php