[gmx-users] Fwd: KALP-15 in DPPC Tutorial Step 0 Segmentation Fault
Steve Vivian
svivian at uwo.ca
Sun Mar 13 01:02:12 CET 2011
On 03/12/2011 02:10 PM, Justin A. Lemkul wrote:
>
>
> Steve Vivian wrote:
>> On 03/12/2011 01:36 PM, Justin A. Lemkul wrote:
>>>
>>>
>>> Steve Vivian wrote:
>>>> On 03/12/2011 12:51 PM, Justin A. Lemkul wrote:
>>>> Getting hard to follow, so I put my new comments in blue.
>>>> Output from Energy Minimization
>>>>
>>>> Steepest Descents converged to Fmax < 1000 in 1814 steps
>>>> Potential Energy = -3.81270276926664e+05
>>>> Maximum force = 9.38712770623942e+02 on atom 2292
>>>> Norm of force = 2.88274347161761e+01
>>>>
>>>>
>>>>
>>>>>
>>>>> Steve Vivian wrote:
>>>>>> Based on a preliminary test using multiple threads, the issue is not
>>>>>> resolved.
>>>>>> This leads me to believe that my Unit Cell is not built properly.
>>>>>>
>>>>>> Below is the procedure used to build the unit cell. I have
>>>>>> reviewed it many
>>>>>> times, but I would appreciate any input regarding potential
>>>>>> improvements,
>>>>>> specifically on the line using trjconv in the EM/Shrink loop.
>>>>>>
>>>>>> Safe up to here, (I hope)...
>>>>>>
>>>>>> cat KALP_newbox.gro dppc128_whole.gro > system.gro
>>>>>>
>>>>>> update minim.mdp
>>>>>> ; Strong position restraints for InflateGRO
>>>>>> #ifdef STRONG_POSRES
>>>>>> #include "strong_posre.itp"
>>>>>> #endif
>>>>>>
>>>>>> Create Strong Position Restraint for protein
>>>>>> genrestr -f KALP_newbox.gro -o strong_posre.itp -fc 100000 100000
>>>>>> 100000
>>>>>>
>>>>>> Scale Lipid positions by a factor of 4
>>>>>> perl inflategro.pl system.gro 4 DPPC 14 system_inflated.gro 5
>>>>>> area.dat
>>>>>>
>>>>>> Begin loop of repeated Energy Minimizations and Shrinking (repeat
>>>>>> loop approximately 25 times until area is approx 71 Ang sq)
>>>>>> Begin LOOP (from n=1 to n = 26)
>>>>>> grompp -f minim.mdp -c systm_inf_n.gro -p topol.top -o em_n.tpr
>>>>>> mdrun -v -deffnm em_n
>>>>>> trjconv -s em_n.tpr -f em_n.gro -o em_n_out.gro -pbc mol -ur
>>>>>> compact
>>>>>> perl inflategro.pl em_n_out.gro 0.95 DPPC 0 sys_shr_1.gro 5
>>>>>
>>>>> One problem here: you start the loop every time with
>>>>> system_inf_n? What is system_inf_n? It seems that you should
>>>>> start one (non-loop) shrink and then process the subsequent
>>>>> shrinking steps from there. At the end of the loop, you write to
>>>>> sys_shr_1.gro, which then never gets used again.
>>>>>
>>>>> -Justin
>>>> Put my comments in blue to make it easier to read.
>>>>
>>>> I apologize for the poor description here. The first loop begins
>>>> with the System_inflate.gro file created in the earlier process.
>>>> The rest of the first loop is shown.
>>>>
>>>> At the beginning of the second loop, the input file is
>>>> system_shrink_1.gro, which was the output file at the end of loop
>>>> 1. Proceed through the loop, updating the n from 1 to 2 in each
>>>> step. Output file is system_shrink_2.gro.
>>>>
>>>> Input file for loop 3 is system_shrink_3.gro, ....
>>>>
>>>> ....Output from loop 26 is system_shrink_26.gro and output on
>>>> screen provides information on updated Lipid area which has
>>>> achieved target value.
>>>>
>>>> Then proceed to add water (after changing vdw radius of C atoms)
>>>> Add ions
>>>> EM again
>>>> Then Equilibrate.
>>>>
>>>>
>>>> Output from Energy Minimization (after addition of water and ions)
>>>>
>>>> Steepest Descents converged to Fmax < 1000 in 1814 steps
>>>> Potential Energy = -3.81270276926664e+05
>>>> Maximum force = 9.38712770623942e+02 on atom 2292
>>>> Norm of force = 2.88274347161761e+01
>>>>
>>>>
>>>
>>> This output indicates to me that the system is completely
>>> reasonable. If you send me your .gro and .top (off-list), I will
>>> try to debug and report back if I can find anything.
>>>
>>> To confirm, you are using Gromacs 4.5.3? What version of FFTW?
>>> What hardware? What compilers were used for the installation?
>>>
>>> -Justin
>>
>>
>> Using Gromacs 4.5.3 and FFTW 3.2.2 both compiled for double precision.
>> It has been a while since I compiled so I'm not exactly sure, but I
>> recall updating gcc and gfortran during the process.
>> Hardware, what type of specifics are you looking for? # of
>> processors and speed?
>> Attached are current Energy Minimized file and topology file.
>>
>
> Hardware and compiler specs are unnecessary. The problem is the
> coordinate file itself. Look at it in your favorite visualization
> program and you will see that the lipid headgroups are completely
> collapsed in on one another. I don't know how or why this happened.
> If you're using the DPPC topology from Tieleman and the .mdp files
> given in the tutorial, I see no reason for this totally unstable
> configuration. I have never used double precision (and it is not
> necessary for standard MD, and only serves to slow your runs down), so
> it would be interesting to see if the problem is a result of double
> vs. single precision. I can successfully re-minimize the coordinate
> file you sent me using single precision:
>
> Steepest Descents converged to Fmax < 1000 in 335 steps
> Potential Energy = -6.3605750e+04
> Maximum force = 7.1865680e+02 on atom 115
> Norm of force = 3.9649860e+01
>
> For now, I would suggest you test the precision issue. Install
> Gromacs again (making note of your compiler versions, because this is
> important to know!) in single precision and try to re-minimize the
> system and proceed. No need to completely re-build it, as you should
> be able to just run a new EM on your existing system like I did.
>
> -Justin
Questions regarding re-compiling FFTW and Gromacs
Can you recommend what compiler versions to use when I re-install
Gromacs/FFTW?
Questions regarding your re-EM of the co-ordinate file that I sent you.
What mdp file did you use to re-minimize?
Could you send me EM file that you have from the above EM, so that I can
visualize it for comparison to what I have been seeing. As I recall,
all of my visualizations have looked similar, so it would be nice to see
what it should look like. You can email that to svivian at uwo.ca, thank you.
Steve
>
>> Steve.
>>
>>
>>>
>>>>
>>>>>
>>>>>> ar_shr1.dat
>>>>>> End LOOP
>>>>>>
>>>>>> Add water Add ions
>>>>>> Re-run EM
>>>>>> Equilibrate (and watch it all explode)
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> -----Original Message-----
>>>>>> From: gmx-users-bounces at gromacs.org
>>>>>> [mailto:gmx-users-bounces at gromacs.org]
>>>>>> On Behalf Of Justin A. Lemkul
>>>>>> Sent: Thursday, March 10, 2011 12:56 PM
>>>>>> To: Discussion list for GROMACS users
>>>>>> Subject: Re: [gmx-users] Fwd: KALP-15 in DPPC Tutorial Step 0
>>>>>> Segmentation
>>>>>> Fault
>>>>>>
>>>>>>
>>>>>>
>>>>>> Steve Vivian wrote:
>>>>>>> On 03/08/2011 10:23 PM, Justin A. Lemkul wrote:
>>>>>>>>
>>>>>>>> Steve Vivian wrote:
>>>>>>>>> New to Gromacs.
>>>>>>>>>
>>>>>>>>> Worked my way through the tutorial with relatively few issues
>>>>>>>>> until the Equilibration stage. My system blows up!!
>>>>>>>>>
>>>>>>>>> Returned to the Topology stage and rebuilt my system ensuring
>>>>>>>>> that I followed the procedure correctly for the InflateGro
>>>>>>>>> process. It appears to be correct, reasonable lipid area, no
>>>>>>>>> water inside my bilayer, vmd shows a structure which appears
>>>>>>>>> normal (although I am new to this). There are voids between
>>>>>>>>> bilayer and water molecules, but this is to be expected, correct?
>>>>>>>>>
>>>>>>>>> Energy Minimization repeatedly produces results within the
>>>>>>>>> expected range.
>>>>>>>>>
>>>>>>>>> Again system blows up at equilibration, step 0 segmentation
>>>>>>>>> fault. Regardless of whether I attempt the NVT or Anneal_Npt
>>>>>>>>> process (using the provided mdp files, including the updates
>>>>>>>>> for restraints on the protein and the lipid molecules).
>>>>>>>>>
>>>>>>>>> I have attempted many variations of the nvt.mdp and
>>>>>>>>> anneal_npt.mdp files hoping to resolve my issue, but with no
>>>>>>>>> success. I will post the log information from the nvt.mdp
>>>>>>>>> file included in the tutorial.
>>>>>>>>>
>>>>>>>>> Started mdrun on node 0 Tue Mar 8 15:42:35 2011
>>>>>>>>>
>>>>>>>>> Step Time Lambda
>>>>>>>>> 0 0.00000 0.00000
>>>>>>>>>
>>>>>>>>> Grid: 9 x 9 x 9 cells
>>>>>>>>> Energies (kJ/mol)
>>>>>>>>> G96Angle Proper Dih. Improper
>>>>>>>>> Dih. LJ-14 Coulomb-14
>>>>>>>>> 8.52380e+01 6.88116e+01 2.23939e+01
>>>>>>>>> -3.03546e+01 2.71260e+03
>>>>>>>>> LJ (SR) Disper. corr. Coulomb
>>>>>>>>> (SR) Coul. recip. Position Rest.
>>>>>>>>> 1.49883e+04 -1.42684e+03 -2.78329e+05
>>>>>>>>> -1.58540e+05 2.57100e+00
>>>>>>>>> Potential Kinetic En. Total
>>>>>>>>> Energy Conserved En. Temperature
>>>>>>>>> -4.20446e+05 *1.41436e+14 1.41436e+14
>>>>>>>>> 1.41436e+14 1.23343e+12*
>>>>>>>>> Pres. DC (bar) Pressure (bar) Constr. rmsd
>>>>>>>>> -1.56331e+02 5.05645e+12 1.18070e+01
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> As you can see the Potential Energy is reasonable, but the
>>>>>>>>> Kinetic Energy and Temperature seem unrealistic.
>>>>>>>>>
>>>>>>>>> I am hoping that this is enough information for a more
>>>>>>>>> experienced Gromacs user to provide guidance. Note: that I
>>>>>>>>> have tried all of the suggestions that I read on the mailing
>>>>>>>>> list and in the "blowing up" section of the manual, specifically:
>>>>>>>>> -reduced time steps in Equilibration Stages
>>>>>>>>> -reduced Fmax during EM stage (down as low as 100kJ which did
>>>>>>>>> not help)
>>>>>>>>> -modified neighbours list parameters
>>>>>>>>>
>>>>>>>>> Any help is appreciated. I can attach and forward any further
>>>>>>>>> information as required, please let me know.
>>>>>>>>>
>>>>>>>> Which Gromacs version are you using? It looks like you're
>>>>>>>> running in serial, is that correct? Otherwise, please provide
>>>>>>>> your mdrun command line. If you're using version 4.5.3 in
>>>>>>>> serial, I have identified a very problematic bug that seems to
>>>>>>>> affect a wide variety of systems that could be related:
>>>>>>>>
>>>>>>> Yes I am currently using Gromacs 4.5.3 in serial.
>>>>>>>
>>>>>>>> http://redmine.gromacs.org/issues/715
>>>>>>>>
>>>>>>>> I have seen even the most robust tutorial systems fail as well,
>>>>>>>> as some new lab members experienced the same problem. The
>>>>>>>> workaround is to run in parallel.
>>>>>>>>
>>>>>>> If I understand you correctly, the recommended workaround is to
>>>>>>> re-configure gromacs 4.5.3 with mpi enabled and complete the
>>>>>>> Equilibration and Production simulation in parallel.
>>>>>>>
>>>>>>
>>>>>> Strictly speaking, an external MPI library is no longer
>>>>>> required. Gromacs
>>>>>> now builds with internal threading support (as long as your
>>>>>> hardware and
>>>>>> compilers support such features). In fact, thread support builds
>>>>>> by default if
>>>>>> possible, so if your mdrun has an -nt flag, you don't need to do
>>>>>> anything else except
>>>>>> use "mdrun -nt (number of threads)" when running your command.
>>>>>>
>>>>>>> Do you have a recommendation for which mpi library to install
>>>>>>> (lam mpi seems to be referenced in other articles on the mailing
>>>>>>> list)?
>>>>>>>
>>>>>>
>>>>>> I've had good luck with OpenMPI in the past, but this is not
>>>>>> strictly
>>>>>> necessary in all cases.
>>>>>>
>>>>>>> Are there documented installation procedures for this process
>>>>>>> (upgrading to gromacs with mpi enabled)?
>>>>>>>
>>>>>>
>>>>>> http://www.gromacs.org/Downloads/Installation_Instructions#Using_MPI
>>>>>>
>>>>>> -Justin
>>>>>>
>>>>>>> Thanks for your assistance.
>>>>>>> Steve.
>>>>>>>
>>>>>>>> -Justin
>>>>>>>>
>>>>>>>>> Regards,
>>>>>>>>> Steve Vivian.
>>>>>>>>> svivian at uwo.ca
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>
More information about the gromacs.org_gmx-users
mailing list