[gmx-users] tpi segmentation fault

[Javier Cerezo]

Thanks a lot for your reply Dr. Hess.

As you expected, the -nt 1 option in the mdrun fix the problem in 
4.5-beta version. Actually, the problem there was not a "segmentation 
fault" warning appearing in the output, it just got stop in an 
apparently random behavior. With 4.0.7 version that I have compiled from 
source I still have the same "Segmentation fault" problem. I tried with 
a trajectory generated with the same compiled version but it still fails.

Anyway, I guess it can be due to a mistake during my compilation since 
this feature had no such failure reports in the mailing list, and in 
preliminary tries, it seems that the results for the tpi calculation in 
4.0.7 and 4.5beta versions are similar. Concretely, I am trying to use a 
modification of the tpi program similar to the slab-tpi proposed by 
Robert Vacha 
(http://lists.gromacs.org/pipermail/gmx-developers/2008-December/002931.html) 
but, for practical reasons the partition of the box is performed 
specifying the number of slices in which the box is partitioned 
(nslices) and in which partition the particle are inserted (islice). For 
that I just modified tpi.c (line 431) to:
    x_init[ZZ] = (gmx_rng_uniform_real(tpi_rand) + islice - 
1)*state->box[ZZ][ZZ]/nslices;

where nslice and islice are correctly passed to the program though 
unused mdp parameters in tpi calculations (nstxtcout and nstenergy). I 
applied it for the study of a lipid bilayer and as a results I expected 
a nice simmetrical curve for <mu> vs. Z-positions (similar to the ones 
in J. Phys. Chem. B 2007, 111, 12748-12755) but I got rough irregular 
curves and partition dependent (the shape depends on nslices). I am 
using pseudo-NPT ensemble (Berendsen coupling) in the production step 
and nsteps=10000 nstlist=100 for tpi insertion. I tried to use different 
versions to investigation why the results are not the expected ones. But 
maybe there is something wrong in the modification (for exmaple, I think 
that there might be some issues since the box size and subsequently the 
partitions are being modified during the trajectory). Does someone have 
any idea about that?

Thank you.

Javier



> Date: Mon, 16 Aug 2010 11:49:21 +0200
> From: Berk Hess<gmx3 at hotmail.com>
> Subject: RE: [gmx-users] tpi segmentation fault
> To: Discussion list for GROMACS users<gmx-users at gromacs.org>
> Message-ID:<COL113-W26902708BA85122A825A338E9B0 at phx.gbl>
> Content-Type: text/plain; charset="iso-8859-1"
>
>
> Hi,
>
> I don't understand why 4.0.7 gives a segv.
>
> But 4.5-beta probably gives a segv because of a parallel problem.
> Could you try 4.5-beta with the mdrun option -nt 1
> and report back?
>
> Thanks,
>
> Berk
>
>    
>> Date: Fri, 13 Aug 2010 18:35:55 +0200
>> From:jcb1 at um.es
>> To:gmx-users at gromacs.org
>> Subject: [gmx-users] tpi segmentation fault
>>
>> Hi all.
>>
>> I am trying to perform a tpi (test particle insertion) calculation on a
>> trajectory generated with mpi_mdrun (gromacs 4.0.7, run in a Beowulf
>> cluster of Intel(R) Core(TM)2 Quad CPU    Q6600  @ 2.40GHz). I am using
>> tpi integrator in the mdp file and the following command:
>>
>>    $  grompp -f tpi.mdp -c 32hoa_128dmpcwrun.gro -n index_tpi.ndx -p
>> topol_tpi.top
>>    $  mdrun -rerun 32hoa_128dmpcwrun.trr -g tpi.log
>>
>> I tried different versions of gromcas and for 4.0.X I got a segmentation
>> fault getting the message:
>>
>>           trn version: GMX_trn_file (single precision)
>>           Reading frame       0 time    0.000   Segmentation fault
>>
>> using a compiled version with "config --prefix
>> /home/cerezo/Programs/gromacs4.07" + "make" + "make install". However,
>> it works when I used a precompiled version (got with apt-get, I have
>> kubuntu). So I guess that it may be due to a mistake during the
>> compilation. In addition, other integrators (i.e. md, steep) work
>> correctly with the version I've compiled. My system is and Intel(R)
>> Core(TM)2 Quad CPU    Q8400  @ 2.66GHz. Is there any trick I could try
>> in the compilation?
>>
>> I also tried gromacs-4.5-betaX and it works but sometimes (apparently
>> randomly) the calculation gets stops (as if it had entered in an
>> infinite loop) after reading the last frame.
>>
>> Thanks for your attention!
>>
>> Javier
>>
>> -- 
>> Javier CEREZO BASTIDA
>> Estudiante de Doctorado
>> ---------------------
>> Dpto. Química-Física
>> Universidad de Murcia
>> 30100 MURCIA (España)
>> Tlf.(+34)868887434
>>
>> -- 
>> gmx-users mailing listgmx-users at gromacs.org
>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>> Please search the archive athttp://www.gromacs.org/search  before posting!
>> Please don't post (un)subscribe requests to the list. Use the
>> www interface or send it togmx-users-request at gromacs.org.
>> Can't post? Readhttp://www.gromacs.org/mailing_lists/users.php
>>      
>   		 	   		
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL:http://lists.gromacs.org/pipermail/gmx-users/attachments/20100816/b7ad9340/attachment-0001.html
>
> ------------------------------
>
>    

-- 
Javier CEREZO BASTIDA
Estudiante de Doctorado
---------------------
Dpto. Química-Física
Universidad de Murcia
30100 MURCIA (España)
Tlf.(+34)868887434