Hey Tom,
we checked the tests you mentioned and it seems that they all pass. I
attached the log-file of the amber parallel tests to this mail. As the
tests seem to be all good, a new compilation of amber and MPI shouldn't
be needed, right?
I have to excuse for a missunderstanding: We WANT to do the calculation
on a single machine, but it DOESN'T WORK, so the answer should have been
NO to this question.
Am 21.06.2012 13:48, schrieb steinbrt.rci.rutgers.edu:
> Hi,
>
>> > Are output files produced? YES
>> > Do they contain error messages? NO
> So output just stops at some point? or mdout is empty? Is the first energy
> summary printed?
>
>> > Did Amber12 pass the parallel test suite? YES
>> > Did other parallel TI jobs run ok before? NOT IN AMBER12
> Hm, did the Amber test suite TI jobs really pass? There is a set of tests
> $AMBERHOME/test/ti* and $AMBERHOME/test/softcore/* that try out the TI-MPI
> calcs. Make sure these work ok before checking your own system.
>
>> > Do you use the right MPI version (the one compiled with Amber) or the
>> pre-installed one?
>> We use Debian 604 with pre-installed open MPI 142
> There can be problems when Amber and MPI are compiled with different
> compilers, I always suggest making a fresh MPI setup with every recompile
> of Amber. But if you say your tests passed, maybe thats not it.
>
>> > Does it run on a single machine (with MPI, but no submission queue
>> etc) YES
> Hm, so you can run the TI calculation on a single machine with MPI, but
> not on your cluster? This sounds like a permission or queueing problem
> then.
>
>> > Can you run single CPU calculations (no TI) with both of your
>> prmtops? YES
>> > Maybe your two endstate systems are not identical in composition?
>> the files are identical, only whith different charge
>> distribution on one HEM-residue.
> ok, if your prmtop files differ only in the list of charges and the restrt
> files are identical, then not much can go wrong code-wise, focus on
> checking if your Amber installation really is fully functional by looking
> at the test results etc.
>
> Thomas
>
> Dr. Thomas Steinbrecher
> formerly at the
> BioMaps Institute
> Rutgers University
> 610 Taylor Rd.
> Piscataway, NJ 08854
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
--
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jun 21 2012 - 06:00:03 PDT
