Hello, I have written a C/C++ code on unstructured mesh (tetrahedra) , using pthreads but not MPI. 1 Is it a good idea to use petsc as a framework as an easier high level library for interface to MPI in order to parallelize the code 2. If yes. I have fields on various location on my mesh ; some variables are defined at cell center, otheres on edges, and others on faces. Does petsc allow to avoid to deal with low level MPI routines to get and set the ghost values between partitions to evaluate some functions such as fluxes at cell faces which needs different variables which may lie on different partitions ? (I have seen on the talk http://www.mcs.anl.gov/petsc/petsc-as/documentation/tutorials/ARSC2010.pdf page 81 some code lines which suggest that it is simple /∗ Code to communicate nonlocal ghost point data ∗/ VecGetArray(F, &f ); /∗ Code to compute local function components ∗/ VecRestoreArray(F, &f ); return 0; Would it be possible to get some advice please ? Thanks Tahar -------------------------------------------- T. Amari Centre de Physique Theorique Ecole Polytechnique 91128 Palaiseau Cedex France tel : 33 1 69 33 42 52 fax: 33 1 69 33 49 49 email: <mailto:[email protected]> URL : http://www.cpht.polytechnique.fr/cpht/amari