I have two PETSc pull requests for VecScatter: PR #1037 adds MPI3 process shared memory, and PR #1047 adds MPI3 neighborhood collective (MPI_Ineighbor_alltoallv). In the master branch, there already is another approach applying MPI3 process shared memory to VecScatter. You can enable it with command line option -vecscatter_type mpi3
--Junchao Zhang On Fri, Sep 7, 2018 at 12:26 PM Tamara Dancheva < [email protected]> wrote: > Hi, > > I am developing an asynchronous method for a FEM solver, and need a custom > implementation of the VecScatterBegin and VecScatterEnd routines. Since > PETSc uses its own limited set of MPI functions, could you tell what would > be the best way possible to extend upon it and use for example the > non-blocking collectives, igatherv and so on? > > I hope the question is specific enough, let me know if anything, I can > provide with more information. I would very much appreciate any help, > thanks in advance! > > Best, > Tamara >
