Re: [AMBER-Developers] pbsa crashes

From: Scott Brozell <sbrozell.rci.rutgers.edu>
Date: Fri, 16 Oct 2009 03:22:43 -0400

Hi,

Testing with
gcc version 4.1.2 20071124 (Red Hat 4.1.2-42)
has all the pbsa tests passing.

scott


On Thu, Oct 15, 2009 at 03:14:46PM -0700, MengJuei Hsieh wrote:
>
> I just committed our current progress on pbsa by cleaning up some
> subroutines with incompatible licensing and adding additional non-linear
> solvers. Let's see if you still get the same errors on this new one.
>
> I am using gfortran (gcc version 4.5.0 20091001) on mac os x 10.6 for my
> primary testing machine. But I am not sure if the gfortran version is a
> deterministic factor.
>
> > From: Andreas Svrcek-Seiler <svrci.tbi.univie.ac.at>
> > the current version of pbsa crashes during various pbsa test-cases,
> > when I compile pbsa with gfortran (I tried 4.3.2 and 4.4.2).
> > Curiously, every 10th time or so I run e.g.
> > $AMBERHOME/test/pbsa_trx/Run.trxox.min, it runs through.
> > However, valgrinding the run
> > (as in valgrind ../../exe/pbsa -O -i mdin -c trxox.2.4ns.x ),
> > I get loads of errors starting with
> > Conditional jump or move depends on uninitialised value(s)
> > at 0x402B42: pbsa_ (_pbsa.f:514)
> > Uninitialised value was created by a stack allocation
> > at 0x402518: pbsa_ (_pbsa.f:165)
> > ...which doesn't tell me much (of course I looked up those lines).
> >
> > When I compile pbsa with ifort, the same test always runs through but
> > valgrind still complains, but curiously at other locations:
> > Conditional jump or move depends on uninitialised value(s)
> > solvent_accessibilitysa_driver_mp_exclud_ (_sa_driver.f:1252)
> > ...
> > Uninitialised value was created by a stack allocation
> > solvent_accessibilitysa_driver_mp_sa_arc_ (_sa_driver.f:970)
> >
> > Conclusion: I don't understand this and was unable to find the "real"
> > cause. Maybe someone else should hav a look. Usually ignoring valgrind
> > warnings comes back at you.
> >
> > best
> > Andreas
> >
> > P.S.: According to an Intel(R)-guy, valgrind gives false warnings
> > when icc/ifort use higly optimized string functions (strlen() etc.

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Oct 16 2009 - 00:30:04 PDT
Custom Search