Error while trying to run lammps on GPU

Hi ,

I compiled lammps with gpu on the cluster in my home directory on a node with gpu.

But I am unable to run lammps on the gpu.

I used cuda-5.5.22 for compilations

I ran ./nvc_get_devices . I got this error :

Cuda driver error 100 in call at file ‘./geryon/nvd_device.h’ in line 267.
Attempting to use an MPI routine before initializing MPICH

when ran ./nvidia-smi to check for gpus , I recieved the following output.

Did you ever get this problem fixed? I assumed

someone else on the mail list might reply, but
I don’t see any responses yet …