[gmx-users] validating gromacs installation 4.0.4

[Y. U. Sasidhar]

Dear Users, I posted the following message a few days ago; So far I have 
not got any response. Searching archives also seem to indicate that 
people are facing similar problems. Some times modifying mdp options 
make the tests "pass" as we also observed ( see below ).

Further the reference trajectories seem to be generated by versions 
3.2/3.3.

Are there reference trajectories for 4.0.4?

If some of the tests "fail", does it mean gromacs is not computing as it 
should? the results can not be relied on?

As of now numbers are compared to test pass/failure.

Can physical properties - like rdf of spc water, diffusion constants 
etc- be used to test gromacs?

I thank you very much for your time.

The "failures" are presenting a dilemma. I hope I have done what I can
as a user and hope to resolve the issue with your help.

regards,

Sasidhar

PS: we notice that with double precision failures are less; further with 
version 3.2 the failures are less suggesting that numerical accuracies 
of machines, differences in algorithms, if any, used in different 
versions could be a cause for "failures"
==============previously posted message=============



Dear users,

We have installed gmx 4.0.4 ( single precision, on cent os 4.3 ( 32 bit
) on quad core dual xeon machine ( clock 2 GHz )

Installation directions as given on the site followed.
And the tests are run using the perl script provided plus test
files/folders.

We find many failures ( see below ); However, by changing mdp options
all the tests "passed" as detailed below. For example, for the "field"
test coulomb type changed to PME and the test passes with this change.

The implication is that reference trajectory computed using cut-off
"matches" with newly calculated trajectory using more accurate PME
method. You can see more details below. **So did the tests really pass?**

We understand energies, virials and forces are compared for all tests to
report errors/failures.


When we searched archives we did not find clear solutions to
test-failure-problems. Wiki site does not give any further guidance.
We have looked at wiki site.

In the presence of failures it is not clear how to proceed further.

Kindly guide us in upgrading to gmx 4.04 and "clearing" the tests.
regards,
Sasidhar





==========================================================================
This is the output (before making any changes), when i get when i run the
test script gmxtest.pl for 4.0.4 on gromacs-4.0.4.

$ ./gmxtest.pl all
All 16 simple tests PASSED
FAILED. Check files in field
FAILED. Check files in tip4p
FAILED. Check files in tip4pflex
FAILED. Check files in water
4 out of 14 complex tests FAILED
FAILED. Check files in kernel020
FAILED. Check files in kernel120
FAILED. Check files in kernel121
FAILED. Check files in kernel122
FAILED. Check files in kernel123
FAILED. Check files in kernel124
FAILED. Check files in kernel220
FAILED. Check files in kernel221
FAILED. Check files in kernel222
FAILED. Check files in kernel223
FAILED. Check files in kernel224
FAILED. Check files in kernel320
FAILED. Check files in kernel321
FAILED. Check files in kernel322
FAILED. Check files in kernel323
FAILED. Check files in kernel324
16 out of 63 kernel tests FAILED
N      Reference   This test
   10    -33.9883    -29.4637
   11    -33.9883    -29.4637
There were 2 differences in final energy with the reference file
All 45 pdb2gmx tests PASSED
pdb2gmx tests FAILED


The tests which failed previously, passed when following changes were made
to the grompp.mdp files of the failed test directories.

1) field        : coulombtype : cut-off to PME
2) tip4p        : tempcoupl   : berendsen to V-rescale
3) tip4pflex    : vdwtype     : cut-off to shift
4) water        : tempcoupl   : yes  to V-rescale

5)kernel020-124 : coulombtype : cut-off to PME & making rlist = rcoulomb
6)kernel220-224 : coulombtype : recation-field-nec to PME & making rlist =
                                    rcoulomb

7)kernel320-324 : coulombtype : switch to PME & making rlist = rcoulomb.