[gmx-users] Problems with GPU-Version and implicit water models

[Rossen Apostolov]

  Hi,


On 08/31/2010 12:53 PM, Christian Mücksch wrote:
>  Dear all,
>
> I am using Gromacs-4.5-beta4 and compiled the GPU-Version. So far only 
> Amber-FF are working in the GPU-version. 
Gromacs-GPU supports only AMBER and CHARMM forcefields. By default, 
pdb2gmx enables the -cmap option, and cmap dihedrals are not supported. 
If you want to use CHARMM, then run as:

$ pdb2gmx -f protein.pdb -nocmap.

(If you use AMBER then -nocmap is not needed of course)

OPLS and Gromos use specific combination rules for nonbonded potential 
parameters and are not supported.

> When I want to use the GBSA implicit water-model I have to use cut-off 
> for the electrostatic interactions due to this error-message:
>

This was a bug and I fixed it for the final release. Thank you for 
reporting!

Rossen

> ERROR 1 [file PR.mdp]:
>   With GBSA, coulombtype must be equal to Cut-off
>
> Now, my problem is that the GPU-version does not accept cut-off for 
> the electrostatic interactions since only options like 
> Reaction-Field,  Ewald,  PME and No-cutoff are supported so I get the 
> following error:
>
> -------------------------------------------------------
> Program mdrun-gpu_beta, VERSION 4.5-beta4
> Source code file: ../gromacs-4.5-beta4/src/kernel/openmm_wrapper.cpp, 
> line: 1038
>
> Fatal error:
> Internal error: you should not see this message, it means that 
> theelectrosatics option check failed. Please report this error!
> For more information and tips for troubleshooting, please check the 
> GROMACS
> website at http://www.gromacs.org/Documentation/Errors
> -------------------------------------------------------
>
> Will there be an update allowing to use implicit-water with 
> reaction-field on the GPU? Will it be possible to use other ff like 
> OPLSAA on the GPU?
>
> Kind regards,
> Christian Muecksch