Hello,
I downloaded this tarball and made 4 different installations to test
different environments (all Linux installations are done on ubuntu 9.04 with
csh, flex, bison, gfortran, etc. installed from software repositories):
1. Linux, intel 11.0.083 compilers, intel MKL 11
2. Linux, intel 11.0.083 compilers, goto
3. Linux, gcc 4.3.3 gfortran 4.3.3, goto
4. Mac OS X 10.6.2, gcc 4.2.1 gfortran 4.5.0, goto
(I do not have intel compilers for the Mac, so I did not run those tests).
I tested ptraj.MPI on the first two using MPICH2-1.1.0
1 and 2 passed all tests with minor deviations in some of the charge
derivations for antechamber, but well within tolerable limits IMO. The only
difference not using MKL was the "global minimum," yet both values were
within the tolerable limits.
3 showed the same variations as 1 and 2, but failed dbf_1.polyALA force and
dbf_1.polyAT force. I removed the "rm" statements in the respective run
scripts and 'diff -y' the resulting file with the save. The first chunk,
Atomic qE Forces, were exactly the same, but Atomic DB Forces were all
considerably different on many occasions (well more than 10% difference in
some occasions).
4. showed the same variations in antechamber as the first two, but failed
much worse in the pbsa_frc section.
checking argasp force
FAILED: argasp force
dbf_1.dadt
checking dadt force
FAILED: dadt force
dbf_1.dgdc
checking dgdc force
FAILED: dgdc force
dbf_1.lysasp
checking lysasp force
FAILED: lysasp force
dbf_1.polyALA
checking polyALA force
FAILED: polyALA force
dbf_1.polyAT
checking polyAT force
FAILED: polyAT force
In addition to these, it also printed to TEST_FAILURES.diff some potentially
worrisome values for EPB. I've included the polyALA.diff, polyAT.diff (test
case 3), and TEST_RESULTS.diff (test case 4) on this email.
On a side note, there has been a decent amount of discussion on the Amber
mailing list with users having trouble compiling and running Amber on
Ubuntu, and since it's a fairly popular distro due to its ease of use it may
be helpful to keep a list of common pitfalls users run into (like not having
csh installed by default) and which packages to install (apt-get) to fix
it. The same could probably be said for Macs, especially with the
architecture problems snow leopard so generously introduced. Since a large
fraction of Amber users likely fall into one of the categories above,
perhaps it would be useful to have some place where developers (and users?)
could list easy fixes to common problems associated with these two platforms
so issues don't continue to periodically circulate on the reflector. What
do others think?
All the best,
Jason
On Mon, Dec 7, 2009 at 9:53 AM, case <case.biomaps.rutgers.edu> wrote:
> I have placed the current release candidate here:
>
> http://ambermd.org/downloads/AmberTools.04dec09.tar.bz2
>
> I think we are getting close! As usual, it is like pulling teeth to get
> anyone to look at and update the documentation. I may have to just lock
> all developers in a room at the January meeting and require a certain
> number
> of updates before allowing them out. But please make changes now if you
> can.
> As I have mentioned before, this is an excellent opportunity for students
> and post-docs to gain some recognition outside their own group by
> contributing
> to a common project.
>
> ...thx...dac
>
>
> _______________________________________________
> 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. Graduate Student
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Dec 07 2009 - 15:00:04 PST
