Re: [AMBER-Developers] NAB and PBSA broken in CVS?

From: Ben Roberts <roberts.qtp.ufl.edu>
Date: Thu, 29 Oct 2009 22:17:51 -0400

Hi,

On Thu, 2009-10-29 at 20:43 -0400, Jason Swails wrote:
> Hey,
>
> The latest version I checked out, 10-22-2009, is passing those tests just
> fine with ifort11 and mkl10 on Linux. The same test fails on Mac with
> ifort11 and mkl11 to the best of my recollection, but Ben can comment
> further on that (he's done all the tests there are to do for Amber on Mac).
> gcc is failing those tests (no MKL) on my Mac as well. My installation with
> PGI 9.0.3 (no Intel MKL) runs just fine and passes those tests (on NICS
> kraken). pbsa test also passed on linux with intel11 and mkl10.
>
> All the best,
> Jason
>
> On Thu, Oct 29, 2009 at 7:31 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
>
> > Hi All,
> >
> > Has anyone else noticed problems running the test cases for AMBER Tools?
> >
> > I just tried the following with ifort l_fc_p_10.1.018 and mkl 10.1.1.019
> >
> > cvs co amber11
> > cd amber11/src/
> > ./configure -static intel
> > make -f Makefile_at
> > cd ../test/
> > make -f Makefile_at
> >
> > and I get a segfault on the very first test case:
> >
> > ( cd nab; make test )
> > make[1]: Entering directory
> > `/server-home/rcw/cvs_checkouts/amber11/test/nab'
> > =====================================================
> > Running test to make dna duplex:
> >
> > /bin/sh: line 1: 12593 Segmentation fault ./duplex <duplex.in
> > make[1]: *** [duplex_test] Error 139
> > make[1]: Leaving directory
> > `/server-home/rcw/cvs_checkouts/amber11/test/nab'
> > make: *** [test.nab] Error 2
> >
> > If I remove test.nab as a target from the default test cases then things
> > work with test failures for the ash and sustiva antechamber tests due to
> > variations in the charges in the 3rd decimal place. I then get a bunch of
> > differences for 1ubqH.pdb in test/reduce.
> >
> > Finally things keep running fine until pbsa_trx where Run.trxox.min dies
> > with a segfault. At this point I haven't tried the rest of the test cases.
> >
> > Do other people see the same problems?
> >
> > It looks like we are not ready to ship AMBERTools yet.
> >
> > All the best
> > Ross


I got those sorts of errors on a Mac. NAB duplex and certain other tests
would fail with segmentation faults. This could be avoided by specifying
"-m32" as a compiler flag. Dave suggested in an email to me that NAB
could be trying to link against a 32-bit library somewhere. Thus,
specifying "-m32" as a compiler flag _may help in your case. In the
longer term, I'd argue for a real fix (as opposed to a mere workaround),
but I'm not sure how to start implementing such a "real fix".

I also found the same thing you report in the Antechamber tests, where
ash, sustiva and guanine would fail. Not always the same tests. The
charges would be off by small amounts, and the exact combination of
problems would seem to depend on the compiler and the choice of
architecture (-m32, -m64 or left blank).

pbsa_trx and the other pbsa tests would fail if built using gcc. If icc
was used, they would pass.

I haven't tried building on a Linux machine for a while, so I can't
speak about the current state of play there.

Ross, I've prepared a comprehensive list of all the tests that were
failing in my builds on a Snow Leopard mac. I could forward the list to
you, if you like. I didn't know how appropriate it would be to send it
to the list at large. Is it something you would find helpful?

Ben



_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Oct 29 2009 - 19:30:03 PDT
Custom Search