[gmx-users] Installation/compilation problem ---v3.3.1

[philippe ayala]

Thanks Mark and Patrick for your prompt feedback.

I did look at hydration energies of small molecules and had no problem
reproducing the literature data.  That being said, i noticed recently that
all was not well in some runs.  I don't think it is a bug.  I am more
inclined to think that it is a compilation problem.

Let me expand a  little bit on the problem i'm experiencing with my
binaries.
The problem is as follows:  using g_analyze_d, the average of block averages
is not the same as the same as the overall average.  Getting a data average
should be easy enough, but it is not the case here.

Here is an example (see Details below):
from 0 to 1000ps ->g_analyze_d average is 1.634425e+03
from 1000 to 2000ps -> g_analyze_d average is 1.634059e+03
from 0 to 2000ps -> g_analyze_d average is 1.637816e+03   ---> a difference
of ~ 3kJ/mol compared with the 1ns blocks
 from 2000 to 4000 ps > g_analyze_d average is 1.637791e+03
from 0 to 4000ps -> g_analyze_d average is 1.620341e+03  --> a difference of
~ 17kJ/mol compared with the 2ns blocks

The difference in averages is about 3 kJ/mol for ~ 1million data points (2
blocks of 1000ps vs 1 block of 2000ps, dt=0.002) and 14-17 kJ/mol for ~2
million data points (2 blocks of 2000ps vs 1 block of 4000ps, dt=0.002 ).
This is much greater than the double counting of the 1000ps or 2000ps data
point would contribute.

Also, if i process xvg files with some of my own scripts/programs (perl or
double-precision fortran), i get:
from 0->4000ps -> average is  1634.2

Also, if i trim the xvg file (using grep '^\..000') to retain data points on
every other 0.1 ps or so, g_analyze_d gives an average of 1.634125e+03 over
the 4000ps run, which looks like the right answer this time.

This is all very perplexing.

Thanks
Philippe

-----Details----

$GROMACS/bin/g_analyze_d -f prod-path.6.coul.l0.5.sol.xvg -e 2000

Read 1 sets of 1000001 points, dt = 0.002

                                      std. dev.    relative deviation of
                       standard       ---------   cumulants from those of
set      average       deviation      sqrt(n-1)   a Gaussian distribition
                                                      cum. 3   cum. 4
SS1   1.637816e+03   3.406546e+01   3.406546e-02      -0.168   -0.012


 $GROMACS/bin/g_analyze_d -f prod-path.6.coul.l0.5.sol.xvg -b 2000 -e 4000

Read 1 sets of 1000001 points, dt = 0.002

                                      std. dev.    relative deviation of
                       standard       ---------   cumulants from those of
set      average       deviation      sqrt(n-1)   a Gaussian distribition
                                                      cum. 3   cum. 4
SS1   1.637791e+03   3.383955e+01   3.383955e-02      -0.154   -0.007

$GROMACS/bin/g_analyze_d -f prod-path.6.coul.l0.5.sol.xvg -e 4000
Read 1 sets of 2000001 points, dt = 0.002

                                      std. dev.    relative deviation of
                       standard       ---------   cumulants from those of
set      average       deviation      sqrt(n-1)   a Gaussian distribition
                                                      cum. 3   cum. 4
SS1   1.620341e+03   3.649526e+01   2.580605e-02       0.676    0.008

---Details end----

On 6/22/07, Mark Abraham <mark.abraham at anu.edu.au> wrote:
>
> > Hi all
> >
> > I have a silly problem and would welcome any suggestions.  At this point
> i
> > suspect that this is a installation or compliation problem.
> >
> > First, let me say that i am a very unsophiticated user of gromacs.
> > Please,
> > bear with me if there is a well known fix for this.
> > I have installed gromacs following the instructions on the website
> > (settings
> > for double precision and mpi).  I started using it and everything seems
> to
> > work pretty well and I have run a number of jobs, mostly FEP's to
> > familiarize myself better with Gromacs.  The results i get are in line
> > with
> > published results.  However, on some runs, the average dg/dl values i
> get
> > from the outputfile or from using g_analyze_d are out-of-line with what
> i
> > can obtain if i process the data myself.   For example, i get a very
> > different average values for 2 block-averages of 1ns each compared to
> the
> > average over the same 2ns.  I've tried a few things and it looks like
> the
> > problem resides with my gromacs binaries.  I can expand on this, if need
> > be.
>
> In the absence of a detailed description of what you're doing, it's going
> to be hard to convince anyone that the problem resides with GROMACS rather
> than your use of it, or your other method you refer to - particularly as
> an inexperienced user. I'd suggest developing scripts that do an entire
> calculation so that you have some basis for recording exactly what you did
> and making it reproducible.
>
> > Before I try reinstalling gromacs, i  would appreciate getting some
> > feedback
> > on the key settings for a porper gromacs installation and how to check
> for
> > potential problems.
> > Also, is there a set of test-jobs i can run to convince myself  that i
> > installed gromacs properly and that everything is working properly?
>
> The installation guide on the website is fine for the vast majority of
> systems. I'm not aware of a test suite per se, because MD is not
> reproducible between slightly different hardware. You should be able to
> reproduce ensemble averages with accuracy, however.
>
> Mark
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20070623/5754b190/attachment.html>