[gmx-users] Waterbox simulation on GPU

[devicerandom]

On 10/01/11 17:10, Christian Mötzing wrote:
> Hello,
>
> I got the waterbox simulation as a simple example to test some additions
> I made in GROMACS. It only contains water molecules, so no fancy
> proteins and stuff.

*What* additions?
Does vanilla Gromacs behave the same?
In both cases, which version?

> On CPU it works fine, on GPU I get these values when
> I extract the potential energy and the density with g_energy.
> E-Pot on CPU is around -37k and density naturally not infinite. What do
> I do wrong?
>
>
>      0.000000  -63232.203125         inf
>      2.000000  -265017.812500         inf
>      4.000000  -277360.406250         inf
>      6.000000  -282572.125000         inf
>      8.000000  -285906.281250         inf
>     10.000000  -288693.312500         inf
>     12.000000  -291001.906250         inf
>     14.000000  -293238.875000         inf
>     16.000000  -294145.437500         inf
>     18.000000  -295828.437500         inf
>     20.000000  -296380.000000         inf
>     22.000000  -297290.656250         inf
>     24.000000  -297524.625000         inf
>     26.000000  -298951.593750         inf
>     28.000000  -298996.750000         inf
>     30.000000  -298484.843750         inf
>     32.000000  -298666.312500         inf
>

If vanilla Gromacs works, I'd say your code :)
If not, then please post details of your system (.mdp etc.) and the 
command lines you use.

m.