Hello I'm trying to write an optimization code that can restart from a written on file optimization iteration. Everything is written with EquationSystems:: write() and the Mesh::write (). I can successfully restart the optimization and the results match with the optimization that wrote the files I restarted with. However, they don't match if I pass the petsc arguments -ksp_type preonly -pc_type lu. It seems as these arguments were ignored when I use EquationSystems:: read() even though the first thing I call in my main() function is LibMeshInit init(argc,argv) (both when I read from file or not). I checked with PetscGetArgs() that the command line arguments were passed successfully even after restarting the optimization. This problem seems to happen only with the adjoint_solve(). I checked the matrix and the RHS of the linear system with and without restart and they match, but the solution of the linear system doesn't match if I add the above mentioned flags. Any clue of what could be wrong? Anything else I should check?
The flags I pass for EquationSystems:: read() are READ_DATA | READ_ADDITIONAL_DATA | READ_HEADER. Thanks Miguel ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
