[gmx-users] Fwd: segmentation fault on gpus

Szilárd Páll pall.szilard at gmail.com
Tue Jun 24 19:42:38 CEST 2014


On Tue, Jun 24, 2014 at 7:18 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
>
> On 6/24/14, 12:14 PM, ram bio wrote:
>>
>> hi Justin,
>>
>> Thanks, that makes sense, however I was able to use the same ligand
>> toplogy
>> files with earlier versions like gromacs 4.5.4 to simulate protein ligand
>> complexes. I can simulate protein-ligand complexes with the current
>> versions also but only for the crystal structures that have ligand bound
>> to
>> it, not the modelled proteins with ligands bound.
>>
>
> So the same ligand works elsewhere, but not in a specific complex?  That
> implicates the starting structure or the quality of the protein structure.
> Other notes below.
>
>
>> and also I would appreciate any comments on the mdp file as below:
>>
>> title                  = Bilayer-500
>> cpp                    = /lib/cpp
>> constraint_algorithm = lincs
>> constraints    = all-bonds; added
>> lincs_iter     = 1; added
>> lincs_order    = 4; added
>> integrator             = md;
>> dt                     = 0.002
>> tinit                  = 0;
>> nsteps                 = 1000000 ; 2 ns
>> nstcomm                = 500
>> nstxout                = 5000
>> nstvout                = 5000
>> nstfout                = 0
>> nstxtcout              = 500
>> xtc_precision          = 1000
>> nstlog                 = 500
>> nstenergy              = 500
>> nstlist                = 50
>
>
> This may work well in terms of GPU performance, but could be problematic if
> there are very sensitive cases.

Why would it be problematic?

>
>> ns_type        = grid; added
>> nstcalclr      = 1
>> ; long range interactions
>> coulombtype            = PME
>> rlist                  = 1.4
>> rcoulomb               = 1.2
>> rvdw                   = 1.2

Setting  rlist != rc doesn't make sense for the Verlet scheme.

>> fourierspacing         = 0.16
>> pme_order              = 4
>> vdw-type               = Cut-off
>> cutoff-scheme          = Verlet
>> Tcoupl                 = v-rescale
>> tau_t                  = 0.1     0.1    0.1  0.1 0.1 0.1  0.1
>> 0.1
>> tc-grps                = Protein POP    SOL  NA  CL LIG SOD CLA
>> ref_t                  = 303     303    303  303 303 303   303 303
>
>
> This tc-grp setup is total nonsense.
>
> http://www.gromacs.org/Documentation/Terminology/Thermostats
>
> -Justin
>
>
>> ; Energy monitoring
>> energygrps             = Protein POP SOL SOD CLA LIG NA CL
>> Pcoupl                 = Berendsen
>> pcoupltype             = semiisotropic
>> tau_p                  = 1.0    1.0
>> compressibility        = 4.5e-5 4.5e-5
>> ref_p                  = 1.0    1.0
>> DispCorr    = no ; added
>> ;generate velocites is on at 293K
>> gen_vel                = yes
>> gen_temp               = 303.0
>> gen_seed               = 478905
>> comm-mode       = Linear
>> comm-grps       = System
>>
>> Thanks,
>> Pramod
>>
>>
>>
>>
>> On Mon, Jun 23, 2014 at 8:03 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>>
>>>
>>>
>>> On 6/23/14, 5:53 PM, ram bio wrote:
>>>
>>>> Dear Gromacs Users,
>>>>
>>>> I have been trying to simulate protein-ligand complex using gromacs
>>>> versions that use verlet cutoff scheme on gpus.
>>>>
>>>> These are some of the issues that i could resolve, and any kind of
>>>> suggestion or help is appreciated.
>>>>
>>>> 1. With gromacs 4.6.5 i could simulate crystal structure from pdb but
>>>> not
>>>> modelled protein with ligand always ended up in segmentation fault. The
>>>> MD
>>>> run during equillibration runs for gew thousands of steps and errors out
>>>> as
>>>> segmentation fault, so i have to rerun using previous cpt file, for a
>>>> run
>>>> of 2ns to complete i have to rerun it for 7-10 times.
>>>>
>>>> So if the the system id not minimized, why would it go into rerun?
>>>>
>>>> 2. I though it could be a bug and used 4.6.3 and 5.0 release, with this
>>>> trial i could equillibrate just modelled protein but not with the
>>>> protein-ligand complex. The same error occurs as i mentioned above with
>>>> protein-ligand complex...
>>>>
>>>> I am attaching my mdp file and the minimized structure that i am using
>>>> for
>>>> equillibration.
>>>>
>>>>
>>> The list doesn't accept attachments.
>>>
>>> It seems clear to me that your ligand topology is not stable.  If the
>>> protein works, but the protein+ligand doesn't, what's changing?  That's
>>> the
>>> likeliest source of your problem.
>>>
>>> -Justin
>>>
>>> --
>>> ==================================================
>>>
>>> Justin A. Lemkul, Ph.D.
>>> Ruth L. Kirschstein NRSA Postdoctoral Fellow
>>>
>>> Department of Pharmaceutical Sciences
>>> School of Pharmacy
>>> Health Sciences Facility II, Room 601
>>> University of Maryland, Baltimore
>>> 20 Penn St.
>>> Baltimore, MD 21201
>>>
>>> jalemkul at outerbanks.umaryland.edu | (410) 706-7441
>>> http://mackerell.umaryland.edu/~jalemkul
>>>
>>> ==================================================
>>> --
>>> Gromacs Users mailing list
>>>
>>> * Please search the archive at http://www.gromacs.org/
>>> Support/Mailing_Lists/GMX-Users_List before posting!
>>>
>>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>
>>> * For (un)subscribe requests visit
>>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
>>> send a mail to gmx-users-request at gromacs.org.
>>>
>
> --
> ==================================================
>
> Justin A. Lemkul, Ph.D.
> Ruth L. Kirschstein NRSA Postdoctoral Fellow
>
> Department of Pharmaceutical Sciences
> School of Pharmacy
> Health Sciences Facility II, Room 601
> University of Maryland, Baltimore
> 20 Penn St.
> Baltimore, MD 21201
>
> jalemkul at outerbanks.umaryland.edu | (410) 706-7441
> http://mackerell.umaryland.edu/~jalemkul
>
> ==================================================
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a
> mail to gmx-users-request at gromacs.org.


More information about the gromacs.org_gmx-users mailing list