[gmx-users] FEP w/ PME inner loops

[Berk Hess]

>From: David <spoel at xray.bmc.uu.se>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: Re: [gmx-users] FEP w/ PME inner loops
>Date: Mon, 08 Aug 2005 21:35:42 +0200
>
>On Mon, 2005-08-08 at 12:54 -0400, David L. Bostick wrote:
> > Hello,
> >
> > Berk instructed me awhile ago that the free-energy innerloops that allow
> > for outputting dV/d(lambda) that includes the PME contribution can be 
>found
> > in the CVS code of 2004/11/01.  I would like to check out this version 
>of
> > the code, but I am only familiar with the regular checkout procedure of 
>the
> > latest CVS code as outlined on the web page.  Could someone instruct me 
>on
> > how to check out a version from a specific date? Forgive me if 
>instructions
> > are actually on the web site ... if they are, I somehow missed them.
>man cvs says:
>cvs co -D 2004-11-01 gmx

Indeed.
I assume everything will work fine.
But I can vaguely remember some compilation problems.
Just try and if it goes wrong, mail again.

>
> >
> > Also, given that these innerloops were discontinued in the code, there 
>must
> > have been some problem with them.  Is there any reason for caution when
> > doing FEP with this version? I would be appreciative of any advice in 
>using
> > this code.
>
>Erik will commit the new loops real soon now, but I don't knwo what was
>wrong with the old ones.

Nothing wrong with the old ones.
The new loops also have 64-bit assembly code and probably some other new 
stuff.
There is such a big number of loops now that a different structure
was required to keep everything managable.

The new FE loops will no longer be generated by the innerloop generator 
program,
but will just be normal code. This making coding, debugging and user editing
much easier.

Berk.