On Nov 28, 2012, at 3:00 PM, Michael Povolotskyi <mpovo...@purdue.edu> wrote:
> Thank you. > Looks like also others treated this parameter erroneously as the absolute > tolerance: > > For example, the Libmesh library: > http://libmesh.sourceforge.net/doxygen/classlibMesh_1_1PetscNonlinearSolver.php#a1d5ed777465c1f08785bc321c84c7c71 > > 00486 // Set the tolerances for the non-linear solver. > 00487 ierr = SNESSetTolerances(_snes, this->absolute_residual_tolerance, > this->relative_residual_tolerance, > 00488 this->absolute_step_tolerance, > this->max_nonlinear_iterations, this->max_function_evaluations); > > Am I right that they have a problem? Based on the name of their variable yes it does appear to be wrong. The reason it is a relative tolerance (and not absolute) is that one is computing x = delta x + x and once delta x is sufficiently smaller than x, delta x + x == x (in floating point) so we don't want to iterate past the point where x is not being changed. Barry > Michael. > > On 11/28/2012 03:55 PM, Barry Smith wrote: >> On Nov 28, 2012, at 2:51 PM, Michael Povolotskyi <mpovo...@purdue.edu> wrote: >> >>> Thank you for the very prompt response, >>> but what is the answer: is it a relative step tolerance or an absolute step >>> tolerance that is set in SNESSetTolerances and checked after each step? >> It is a relative test: snorm < snes->stol*xnorm || step || < stol >> || x || where x is the current solution >> >> >> Barry >> >>> Michael. >>> >>> On 11/28/2012 03:47 PM, Barry Smith wrote: >>>> From SNESDefaultConverged() >>>> >>>> else if (snorm < snes->stol*xnorm) { >>>> ierr = PetscInfo3(snes,"Converged due to small update length: >>>> %14.12e < %14.12e * >>>> %14.12e\n",(double)snorm,(double)snes->stol,(double)xnorm);CHKERRQ(ierr); >>>> *reason = SNES_CONVERGED_SNORM_RELATIVE; >>>> } >>>> >>>> I will fix the documentation >>>> >>>> On Nov 28, 2012, at 2:44 PM, Michael Povolotskyi <mpovo...@purdue.edu> >>>> wrote: >>>> >>>>> Hello Petsc developers, >>>>> I have a question about setting tolerances for SNES solver. I have found >>>>> 3 pieces of information that seem to contradict to each other: >>>>> >>>>> 1) >>>>> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/SNES/SNESSetTolerances.html >>>>> stol - convergence tolerance in terms of the norm of the >>>>> change in the solution between steps >>>>> >>>>> 2) >>>>> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/SNES/SNESConvergedReason.html >>>>> SNES_CONVERGED_SNORM_RELATIVE = 4, /* Newton computed step size >>>>> small; || delta x || < stol */ >>>>> >>>>> 3) >>>>> http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/SNES/SNESDefaultConverged.html >>>>> >>>>> <http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/SNES/SNES_CONVERGED_SNORM_RELATIVE.html#SNES_CONVERGED_SNORM_RELATIVE> >>>>> SNES_CONVERGED_SNORM_RELATIVE - (snorm < stol*xnorm) >>>>> >>>>> My question is: >>>>> which criteria is used SNES solver of LS type: absolute, like in 1) and >>>>> 2), or relative, like in 3) ? >>>>> Thank you, >>>>> Michael. >>>>> >>>>> >>>>> >>> >>> -- >>> Michael Povolotskyi, PhD >>> Research Assistant Professor >>> Network for Computational Nanotechnology >>> 207 S Martin Jischke Drive >>> Purdue University, DLR, room 441-10 >>> West Lafayette, Indiana 47907 >>> >>> phone: +1-765-494-9396 >>> fax: +1-765-496-6026 >>> > > > -- > Michael Povolotskyi, PhD > Research Assistant Professor > Network for Computational Nanotechnology > 207 S Martin Jischke Drive > Purdue University, DLR, room 441-10 > West Lafayette, Indiana 47907 > > phone: +1-765-494-9396 > fax: +1-765-496-6026 > ------------------------------------------------------------------------------ Keep yourself connected to Go Parallel: INSIGHTS What's next for parallel hardware, programming and related areas? Interviews and blogs by thought leaders keep you ahead of the curve. http://goparallel.sourceforge.net _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
