Run CUDA programs using MPI in cluster

I want to setup a simple CUDA distributed system using two machine with CentOS 6.3 x86_64.
I have installed CUDA 5.0 and OpenMPI 1.6.2 64 bit in these two machines. How i can run a CUDA program using Open MPI in these two machines.

Please help. I am very new to the LINUX using.

What are the basic steps done for Compiling, building and running CUDA programs using OpenMPI in two machines.

You could learn it from CUDA sample simpleMPI. Good luck!