Hi Dave,
With regard to:
"This is known behavior: pmemd.amoeba will fail badly if there is not enough
stack space, but Bob never figured out how to correctly test for this. At
least the error tells you not to run pmemd.amoeba....We should probably add
a better explanation to the test script." -
Okay, there is the standard attempt to unlimit the stack in the main routine of both pmemd and pmemd.amba (ie., code known to work on most systems at the time of development is in place). Are we hitting variation in system call support for this feature, or what? I actually have no recollection of anyone raising this issue with me, but it would be system-specific (and user privileges-specific) and a bit hard to test. I thought we did rather clearly, at least in the past, both check the stack limits and document the problem for the main pmemd code; the amoeba code has the same checks, but would just be more prone to problems due to massive data use. Still, for these small test cases, no problem should occur. I looked at the Amber 10 code to confirm existence of the check in both pmemd and pmemd.amoeba; I rather doubt that anyone would have removed it in Amber 11 or 12. Could be that the getrlimit() and/or setrlimit() system calls are not working correctly in some linux version, or the spec has changed s
ince I wrote the code.
Regards - Bob
________________________________________
From: David A Case [case.biomaps.rutgers.edu]
Sent: Friday, March 30, 2012 9:14 AM
To: AMBER Developers Mailing List
Subject: Re: [AMBER-Developers] Third (final?) release candidate for Amber 12
On Fri, Mar 30, 2012, steinbrt.rci.rutgers.edu wrote:
>
> Test AT serial: 591 passed, 1 error in sqm/fecp2,
>
> > tail fecp2.pm6.sp.out
> QMMM: System specified with odd number of electrons ( 57)
> QMMM: but odd spin ( 1). You most likely have the charge of
> QMMM: QM region (qmcharge) set incorrectly. Correct error and re-run
> calculation.
>
> then test AT parallel: 37 passed, 0 fail
weird....can you post the entire output file? This one has not been reported
before (should find 58 electrons).
>
> -------------------------------------------------------------------
> - Second unpacked Amber12 on top of that, reconfigured and built:
>
> Test AT serial: 601 passed, same 1 error as above
> Test Amber serial: 481 passed, 6 errors:
>
> in qmmm_EXTERN/pure_QM_MD_GAUSSIAN the scripts:
>
> ./Run.h2o.blyp_sto-3g
> ./Run.h2o.blyp_sto-3g_md
> ./Run.h2o.template.blyp_sto-3g
>
> and in qmmm_EXTERN/QMMM_MD_GAUSSIAN the scripts
>
> ./Run.aladip.hf_sto-3g
> ./Run.nma-spcfw-15.hf_sto-3g
>
> fail because there is no g09 executable. I have g03 but not g09 installed,
> I'm not sure why it fails instead of skipping the test.
Is g03 in your path? Is g03 in $GAUSS_EXEDIR? If you unset your environment
variable GAUSS_EXEDIR, it should skip the tests. On the other hand, it should
work with either g03 or g09. Look at $AMBERHOME/test/check_GAUSSIAN.x.
>
> also in qmmm_EXTERN/QMMM_MD_Orca ./Run.aladip.hf_sto-3g.MechEm fails and
> in qmmm_EXTERN/QMMM_MD_TeraChem ./Run.aladip.hf_sto-3g.MechEm and
> ./Run.nma-spcfw-15.hf_sto-3g.MechEm fail, all three by virtue of not
> existing.
Can you try to see why the check_Orca.x script is not correctly identifying
that you don't have ORCA installed?
>
> Furthermore in amoeba_wat2 test ./Run.ntpverlet.pmemd segfaulted with no
> output but the 'Amber 11 SANDER...' output file header. huh?
This is known behavior: pmemd.amoeba will fail badly if there is not enough
stack space, but Bob never figured out how to correctly test for this. At
least the error tells you not to run pmemd.amoeba....We should probably add
a better explanation to the test script.
...thanks for the report!....dac
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Mar 30 2012 - 10:00:03 PDT
