Pass rank to cuda_init, remove need of including mpi.h there
Pass rank to cuda_init, remove need of including mpi.h there. With this, one does not need a C++ compiler witha MPI wrapper for compiling the Cuda files.
Pass rank to cuda_init, remove need of including mpi.h there. With this, one does not need a C++ compiler witha MPI wrapper for compiling the Cuda files.