[petsc-users] Solutions different after program restart

[Niceno Bojan]

Dear Jed,


> > - First run: 1.98077e-02
> > - Second run: 2.61329e-02
> >
> > The tolerances I request from KSP's gmres solver are:
> >
> > - Linear solver relative tolerance = 0.001;
> > - Linear solver absolute tolerance = 1.0e-5;

> Is the initial residual about 30?

Well, even above that, around 50 :-)

You were right.  Reducing 50 by a factor of 0.001, which I set, the differences are indeed smaller than the solver tolerance :-)


But, allow me to ask one more time, to make sure:  I am using gmres with Jacobi's preconditioner.  The run is sequential, the nodes are not renumbered.  Is it safe to neglect these differences, even if they are smaller than prescribed tolerances?


    Kind regards,


    Bojan



> > For some codes, the domain decomposition will not be identical on
> restart;
> > that changes the algorithm.
>
> That is clear, but what I am comparing are two sequential runs.
>

In that case, check that the system being solved is identical. You can save
it with -ksp_view_binary. Load it with
src/ksp/ksp/examples/tutorials/ex10.c as an independent test. Also be sure
to use the same initial guess.