[gmx-users] RE: Patch for PME in Free Energy?

[van Bemmelen]

Hi Justin,

I can understand that you want your "forward" and "reverse" direction
results to be equal for verification reasons. But keep in mind that
there really is no "forward" or "reverse" direction when doing TI on
separately equilibrated (i.e. independent for every lambda) starting
stuctures (which is IMHO the right way do do TI, btw). There is no
physical reason to check whether both directions give equal results; a
simulation for lambda=0.35 in the forward direction is (or should be)
exactly equivalent to a simulation for lambda=0.65 in the reverse
direction, except for the sign of d(lambda). There is only a
computational reason, obviously, since it helps finding code bugs. ;-)

So if the "forward" direction is working fine (which I believe it is),
than you don't really need this patch. Simply make sure that your
topologies are consistently defined in that direction.

Having said that, and taking my first sentence into account, you could
try downloading the 3.3 CVS version. From the information Berk has given
in bug 175, I'm guessing the code was changed somewhere in pme.c,
nb_free_energy.c or force.c. So you may want to 'diff' those with the
3.3.1 files and alter them accordingly. But since I'm not an expert on
the code, I could be wrong about the files.

Cheers,
Jeroen


>Date: Sun, 17 Feb 2008 17:12:57 -0500
>From: "Justin A. Lemkul" <jalemkul at vt.edu>
>Subject: [gmx-users] Patch for PME in Free Energy?
>To: gmx-users at gromacs.org
>Message-ID: <1203286377.47b8b1694675b at webmail.vt.edu>
>Content-Type: text/plain; charset=ISO-8859-1
>
>Hi all,
>
>I guess this is directed towards the developers.  Thanks to 
>David Mobley, I was
>able to determine that my free energy calculations were victim 
>of a PME bug in
>version 3.3.1.  In the Bugzilla, Berk commented that the issue had been
>resolved for the upcoming release of versions 3.3.3 and 4.0, 
>so I'm guessing
>the issue remains in version 3.3.2?  If so, is there a 
>modification I can make
>to any source code files (either version 3.3.1 or 3.3.2) so I 
>can re-compile,
>or should I just be patient and wait for version 3.3.3?
>
>Thanks,
>Justin