Re: [AMBER-Developers] how to test the AmberTools rc

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 12 Apr 2011 19:24:10 -0700

Sorry -- [tab-space] accidentally sent the unfinished email. I'll finish
now...


> I don't know if I've reported this at all, but I've tried this combination
> with every release candidate that you've made to make sure that it works.
> I've tested it with
>
> Mac OS X 10.6 + GNU 4.4 [ ./configure -macAccelerate {-mpi} gnu ]
> Ubuntu Linux + Intel 11.1.069 (+CUDA) [ ./configure {-mpi &| -cuda} intel ]
> Athena Cray XT 4 + PGI 10.6 [ ./configure -crayxt5 {-mpi} pgi ]
>
> and it compiles successfully. I've always unpacked a fresh amber11 copy
> and applied the bug fixes to AT 1.5 + Amber 11. Some of the patches fail,
> but that's because some of the CUDA patches had to modify configure, which
> is part of AmberTools. These patches just miss completely, though, and
> create a configure.rej and configure.orig, leaving configure alone. This
> may panic some users, but I don't see how that's avoidable.
>
> Also, many of the sander+pbsa tests will fail with numerical differences
> due to changes in PBSA, but I don't see how that's avoidable either (and
> that's beyond the ability of AT15_Amber11.py to deal with unless you want to
> just turn those tests off completely).
>
> The most serious issues come from several PBSA tests that result from (I
> think) a removed namelist variable (I'm not sure which namelist; &cntrl or
> &pb):
>
> cd sander_pbsa_frc && ./test
> dbf_0.argasp
> forrtl: severe (19): invalid reference to variable in NAMELIST input, unit
> 5, file
> /home/swails/newamber/fusion/amber11/test/sander_pbsa_frc/dbf_0/argasp/
> min.in, line 10, position 29
> Image PC Routine Line
> Source
> libintlc.so.5 00002AD3E34A6B1D Unknown Unknown Unknown
> libintlc.so.5 00002AD3E34A5625 Unknown Unknown Unknown
> libifcore.so.5 00002AD3E2CF4D4C Unknown Unknown Unknown
> ...
>
> (All of the tests in this directory failed; these I can turn off in
> AT15_Amber11.py if that's desired.)
>
> Conspicuously absent from my list of tests: sander.PUPIL and sander.APBS.
> However, sander.APBS works just fine with Amber11 and the git tree as of a
> couple months ago, so I don't expect any issues on that front. Also, I've
> never tested trying to extract AmberTools 1.5 on top of a dirty AmberTools
> 1.4 + Amber 11 directory. I *think* for the most part that AmberTools 1.5
> will just overwrite all of the files that are actually used in AmberTools
> 1.4, so it'll work just fine, but I can't say that with 100% certainty.
>

A final note on AT15_Amber11.py. The *most* work it does is on the
config.h, since that file significantly changed (and is no longer
compatible). It also changes sander's and pmemd's Makefiles to adjust for
changes to where netcdf.mod is located. If these lines are changed by
future patches, it won't match the pattern and will just skip it (so patches
can be made against the standard amber11-with-patches and at15-with-patches
branches). It also gets rid of mmpbsa_py source/tests from the Amber
directory (since it's now in AmberTools).

Thus ends my endless treatise,
Jason


-- 
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Apr 12 2011 - 19:30:02 PDT
Custom Search