[gmx-users] gromacs installation and test set

[Justin A. Lemkul]

Gu, Xiang wrote:
> Dear Justin,
> 
> (1) sorry, my ignorance; threads was enabled when installing FFTW 
> instead of Gromacs, I should have said "configured with --enable-float 
> and --enable-mpi" (and actually this is what I did).
> 
> (2) I think I used OpenMPI.
> 

The use of "think" does not inspire confidence.  Knowing your implementation 
(including its version) will help decipher if this is a bug in MPI, since some 
versions have known issues.

> (3) Thank you. Is there anyway to fix the grompp?
> 

I doubt grompp is broken.  Did you compile the entire Gromacs suite with 
--enable-mpi?  If so, you might be getting weird behavior, since mdrun is the 
only program that is MPI-aware.  Otherwise, I'd suspect that the gmxtest 
directory is in the wrong place.  It should be a subdirectory of the "gromacs" 
directory of your installation.  If in the default /usr/local/gromacs, then you 
should have /usr/local/gromacs/gmxtest.

-Justin

> Xiang
> 
> Justin A. Lemkul wrote:
>>
>>
>> Gu, Xiang wrote:
>>> Hi, all,
>>>
>>> I'm trying to install a parallel version of gromacs-4.0.7 under my 
>>> own account (on our campus supercomputing center, configured with 
>>> --enable-threads, --enable-float, --enable-mpi, and prefixed with 
>>> specified directory), because I need to develop some new extensions 
>>> for our research and will have to do code modification and debugging 
>>> from time to time.
>>>
>>
>> There is no point to using --enable-threads in version 4.0.7 - there's 
>> no thread support yet.
>>
>>> Now the installation went through (I felt there was no severe error 
>>> information shown). But when I started running it parallely, the 
>>> mdrun program seemed to crash at the first MPI command it encountered 
>>> (the MPI_Bcast in line 372 of gmxlib/network.c). It is not yet 
>>> addressed what has gone wrong in the installation and I was trying to 
>>> run the test set to get some information. I wonder if I had put the 
>>> gmxtest directory at proper place, since after I sourced GMXRC and 
>>> ran gmxtest.pl, for all the cases I was told "No topol.tpr file in 
>>> kernel ...".
>>>
>>
>> What is your MPI implementation?  OpenMPI, MPICH, etc?
>>
>>> It would be quite appreciated if someone could guess what might have 
>>> been wrong with the installation; also is it worth running the test 
>>> set for locating the problem? and what does the error information "No 
>>> topol.tpr file in kernel ..." imply?
>>>
>>
>> The test set is not terribly useful, and certainly won't diagnose this 
>> issue for you.  If you've failed to generate the necessary .tpr file, 
>> then errors occurred in grompp that prevented the .tpr file from being 
>> generated.
>>
>> -Justin
>>
>>> Thank you!
>>>
>>> Xiang Gu
>>>
>>
> 
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================