[gmx-users] Error while launching kernel pme_solve_kernel: invalid argument

[Gschwend Grégoire]

Hello everyone,


I have the following error message when I try to run a simulation with walls, 2D periodic boundary conditions and 3dc Ewald geometry with Gromacs 2018.2.


"Error while launching kernel pme_solve_kernel: invalid argument"


This error only appears when I try to use a separate pme rank running on GPU. The command is the following:


" gmx mdrun -v -deffnm input -ntmpi 16 -ntomp 2 -npme 1 -pme gpu -nb gpu"


Without a separate pme rank on GPU the simulation works fine, much slower however.


Is this a bug or a limitation of Gromacs with the 3dc Ewald ?


Thank you for your help and best regards !
--
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-***@gromacs.org.

[Mark Abraham]

Hi,

That could well be a bug. If things are not supported then we intend to
give a clear report about what and why. Can you please open an issue at
https://redmine.gromacs.org and attach at least your .tpr (and preferably
grompp inputs).

Mark

On Tue, Nov 27, 2018 at 10:05 AM Gschwend Grégoire <
***@epfl.ch> wrote:

> Hello everyone,
>
>
> I have the following error message when I try to run a simulation with
> walls, 2D periodic boundary conditions and 3dc Ewald geometry with Gromacs
> 2018.2.
>
>
> "Error while launching kernel pme_solve_kernel: invalid argument"
>
>
> This error only appears when I try to use a separate pme rank running on
> GPU. The command is the following:
>
>
> " gmx mdrun -v -deffnm input -ntmpi 16 -ntomp 2 -npme 1 -pme gpu -nb gpu"
>
>
> Without a separate pme rank on GPU the simulation works fine, much slower
> however.
>
>
> Is this a bug or a limitation of Gromacs with the 3dc Ewald ?
>
>
> Thank you for your help and best regards !
> --
> 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-***@gromacs.org.
>
--
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-requ

[Gschwend Grégoire]

Ok, thank you for you help. I just opened an issue on https://redmine.gromacs.org, with a description of the error and the .mdp and .tpr files.


It seems that the error appears with large systems, I will work with a smaller box for the moment...

________________________________
De : gromacs.org_gmx-users-***@maillist.sys.kth.se <gromacs.org_gmx-users-***@maillist.sys.kth.se> de la part de Mark Abraham <***@gmail.com>
Envoyé : mardi 27 novembre 2018 11:37:49
À : gmx-***@gromacs.org
Cc : gromacs.org_gmx-***@maillist.sys.kth.se
Objet : Re: [gmx-users] Error while launching kernel pme_solve_kernel: invalid argument

Hi,

That could well be a bug. If things are not supported then we intend to
give a clear report about what and why. Can you please open an issue at
https://redmine.gromacs.org and attach at least your .tpr (and preferably
grompp inputs).

Mark

On Tue, Nov 27, 2018 at 10:05 AM Gschwend Grégoire <
***@epfl.ch> wrote:

> Hello everyone,
>
>
> I have the following error message when I try to run a simulation with
> walls, 2D periodic boundary conditions and 3dc Ewald geometry with Gromacs
> 2018.2.
>
>
> "Error while launching kernel pme_solve_kernel: invalid argument"
>
>
> This error only appears when I try to use a separate pme rank running on
> GPU. The command is the following:
>
>
> " gmx mdrun -v -deffnm input -ntmpi 16 -ntomp 2 -npme 1 -pme gpu -nb gpu"
>
>
> Without a separate pme rank on GPU the simulation works fine, much slower
> however.
>
>
> Is this a bug or a limitation of Gromacs with the 3dc Ewald ?
>
>
> Thank you for your help and best regards !
> --
> 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-***@gromacs.org.
>
--
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-***@gromacs.org.
--
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-***@gromacs.org.