Another thing to try:
Run "make test.sander.TI" with DO_PARALLEL set to test 2, 4, 16, and 32
processors. When I do this on my MacBook, all tests pass except the
complex-ligand system (test/softcore/complex) and the restrained complex
(test/softcore/complex_rst). A quick look at the failed output files shows
that the temperature of the TI region is several hundred kelvin different
(in addition to differences everywhere else, too).
I don't know if this is related, but it's probably a good place to start.
HTH,
Jason
On Mon, May 23, 2011 at 9:42 PM, Jodi Ann Hadden <jodih.uga.edu> wrote:
> I have come across an interesting problem with SHAKE in TI in Amber 11. I
> have been running tests using two protein-carbohydrate complex systems. My
> calculations crash immediately with the following error:
>
>
> Coordinate resetting (SHAKE) cannot be accomplished,
> deviation is too large
> NITER, NIT, LL, I and J are : 0 0 1014 2057 2058
>
> Note: This is usually a symptom of some deeper
> problem with the energetics of the system.
>
>
> If ALL of the following are true:
>
> 1) The protein is present in the system (regardless of whether the mutation
> is occurring in the protein or in the bound ligand)
> 2) SHAKE is on
> 3) The total number of cpus the job is submitted to is >16
>
> BUT
>
> The same calculation runs without complaint if ONE of the following is
> true:
>
> 1) Only the carbohydrate ligand is present (i.e. Ti for the mutation in the
> free ligand)
> 2) SHAKE is turned off
> 3) The total number of cpus the job is submitted to is 16 or less
>
> So basically, TI with a protein in the system will not run with SHAKE on if
> the job is submitted to >16 cpus. But it will run if I either turn SHAKE off
> or submit to16 or fewer cpus. A system containing only a carbohydrate in
> water runs with SHAKE on regardless of the number of cpus the job is
> submitted to, even if I solvate with a large enough box to give it a greater
> number of atoms than the protein-carboydrate complex system.
>
> I have tested this on multiple machines with different architectures all
> running the release version of Amber 11 (fully patched), as well as the git
> version of AMBER from earlier today, and have gotten the same behavior. In
> case it could provide useful information, I tested for the non-power-of-2
> number of cpus at which the run/crash threshold lies: 18 the protein system
> runs with SHAKE on; 20 it crashes.
>
> Any thoughts?
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon May 23 2011 - 20:30:03 PDT
