[Resa]

I'm trying to solve the flow field in a complex geometry using MPI parallelization. I already worte my own lattice boltzmann code and it's working in both 2d and 3d test cases in series version. Now my question is which technique is the best for domain decomposition. In my previous work using finite element, I used METIS for domain decomposition and then each processor had its own vector of elements and processes its own data. I don't know whether I should make a vector of lattice and then devide it base on the number of processors or should I follow a different strategy. Looking forward to hearing your answers.