<div dir="ltr">Barry, this is probably the Chebyshev problem.<div><br></div><div>Sanjay, this is fixed but has not yet been moved to the master branch. You can fix this now with with -ksp_chebyshev_esteig_random. This should recover v3.5 semantics.</div><div><br></div><div>Mark </div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, May 19, 2016 at 2:42 PM, Barry Smith <span dir="ltr"><<a href="mailto:bsmith@mcs.anl.gov" target="_blank">bsmith@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
We see this occasionally, there is nothing in the definition of GAMG that guarantees a positive definite preconditioner even if the operator was positive definite so we don't think this is a bug in the code. We've found using a slightly stronger smoother, like one more smoothing step seems to remove the problem.<br>
<br>
Barry<br>
<div><div><br>
> On May 19, 2016, at 1:07 PM, Sanjay Govindjee <<a href="mailto:s_g@berkeley.edu" target="_blank">s_g@berkeley.edu</a>> wrote:<br>
><br>
> I am trying to solve a very ordinary nonlinear elasticity problem<br>
> using -ksp_type cg -pc_type gamg in PETSc 3.7.0, which worked fine<br>
> in PETSc 3.5.3.<br>
><br>
> The problem I am seeing is on my first Newton iteration, the Ax=b<br>
> solve returns with and Indefinite Preconditioner error (KSPGetConvergedReason == -8):<br>
> (log_view.txt output also attached)<br>
><br>
> 0 KSP Residual norm 8.411630828687e-02<br>
> 1 KSP Residual norm 2.852209578900e-02<br>
> NO CONVERGENCE REASON: Indefinite Preconditioner<br>
> NO CONVERGENCE REASON: Indefinite Preconditioner<br>
><br>
> On the next and subsequent Newton iterations, I see perfectly normal<br>
> behavior and the problem converges quadratically. The results look fine.<br>
><br>
> I tried the same problem with -pc_type jacobi as well as super-lu, and mumps<br>
> and they all work without complaint.<br>
><br>
> My run line for GAMG is:<br>
> -ksp_type cg -ksp_monitor -log_view -pc_type gamg -pc_gamg_type agg -pc_gamg_agg_nsmooths 1 -options_left<br>
><br>
> The code flow looks like:<br>
><br>
> ! If no matrix allocation yet<br>
> if(Kmat.eq.0) then<br>
> call MatCreate(PETSC_COMM_WORLD,Kmat,ierr)<br>
> call MatSetSizes(Kmat,numpeq,numpeq,PETSC_DETERMINE,PETSC_DETERMINE,ierr)<br>
> call MatSetBlockSize(Kmat,nsbk,ierr)<br>
> call MatSetFromOptions(Kmat, ierr)<br>
> call MatSetType(Kmat,MATAIJ,ierr)<br>
> call MatMPIAIJSetPreallocation(Kmat,PETSC_NULL_INTEGER,mr(np(246)),PETSC_NULL_INTEGER,mr(np(247)),ierr)<br>
> call MatSeqAIJSetPreallocation(Kmat,PETSC_NULL_INTEGER,mr(np(246)),ierr)<br>
> endif<br>
><br>
> call MatZeroEntries(Kmat,ierr)<br>
><br>
> ! Code to set values in matrix<br>
><br>
> call MatAssemblyBegin(Kmat, MAT_FINAL_ASSEMBLY, ierr)<br>
> call MatAssemblyEnd(Kmat, MAT_FINAL_ASSEMBLY, ierr)<br>
> call MatSetOption(Kmat,MAT_NEW_NONZERO_LOCATIONS,PETSC_TRUE,ierr)<br>
><br>
> ! If no rhs allocation yet<br>
> if(rhs.eq.0) then<br>
> call VecCreate (PETSC_COMM_WORLD, rhs, ierr)<br>
> call VecSetSizes (rhs, numpeq, PETSC_DECIDE, ierr)<br>
> call VecSetFromOptions(rhs, ierr)<br>
> endif<br>
><br>
> ! Code to set values in RHS<br>
><br>
> call VecAssemblyBegin(rhs, ierr)<br>
> call VecAssemblyEnd(rhs, ierr)<br>
><br>
> if(kspsol_exists) then<br>
> call KSPDestroy(kspsol,ierr)<br>
> endif<br>
><br>
> call KSPCreate(PETSC_COMM_WORLD, kspsol ,ierr)<br>
> call KSPSetOperators(kspsol, Kmat, Kmat, ierr)<br>
> call KSPSetFromOptions(kspsol,ierr)<br>
> call KSPGetPC(kspsol, pc , ierr)<br>
><br>
> call PCSetCoordinates(pc,ndm,numpn,hr(np(43)),ierr)<br>
><br>
> call KSPSolve(kspsol, rhs, sol, ierr)<br>
> call KSPGetConvergedReason(kspsol,reason,ierr)<br>
><br>
> ! update solution, go back to the top<br>
><br>
> reason is coming back as -8 on my first Ax=b solve and 2 or 3 after that<br>
> (with gamg). With the other solvers it is coming back as 2 or 3 for<br>
> iterative options and 4 if I use one of the direct solvers.<br>
><br>
> Any ideas on what is causing the Indefinite PC on the first iteration with GAMG?<br>
><br>
> Thanks in advance,<br>
> -sanjay<br>
><br>
> --<br>
> -----------------------------------------------<br>
> Sanjay Govindjee, PhD, PE<br>
> Professor of Civil Engineering<br>
><br>
> 779 Davis Hall<br>
> University of California<br>
> Berkeley, CA 94720-1710<br>
><br>
> Voice: <a href="tel:%2B1%20510%20642%206060" value="+15106426060" target="_blank">+1 510 642 6060</a><br>
> FAX: <a href="tel:%2B1%20510%20643%205264" value="+15106435264" target="_blank">+1 510 643 5264</a><br>
><br>
> <a href="mailto:s_g@berkeley.edu" target="_blank">s_g@berkeley.edu</a><br>
> <a href="http://www.ce.berkeley.edu/~sanjay" rel="noreferrer" target="_blank">http://www.ce.berkeley.edu/~sanjay</a><br>
><br>
> -----------------------------------------------<br>
><br>
> Books:<br>
><br>
> Engineering Mechanics of Deformable<br>
> Solids: A Presentation with Exercises<br>
><br>
> <a href="http://www.oup.com/us/catalog/general/subject/Physics/MaterialsScience/?view=usa&ci=9780199651641" rel="noreferrer" target="_blank">http://www.oup.com/us/catalog/general/subject/Physics/MaterialsScience/?view=usa&ci=9780199651641</a><br>
> <a href="http://ukcatalogue.oup.com/product/9780199651641.do" rel="noreferrer" target="_blank">http://ukcatalogue.oup.com/product/9780199651641.do</a><br>
> <a href="http://amzn.com/0199651647" rel="noreferrer" target="_blank">http://amzn.com/0199651647</a><br>
><br>
><br>
> Engineering Mechanics 3 (Dynamics) 2nd Edition<br>
><br>
> <a href="http://www.springer.com/978-3-642-53711-0" rel="noreferrer" target="_blank">http://www.springer.com/978-3-642-53711-0</a><br>
> <a href="http://amzn.com/3642537111" rel="noreferrer" target="_blank">http://amzn.com/3642537111</a><br>
><br>
><br>
> Engineering Mechanics 3, Supplementary Problems: Dynamics<br>
><br>
> <a href="http://www.amzn.com/B00SOXN8JU" rel="noreferrer" target="_blank">http://www.amzn.com/B00SOXN8JU</a><br>
><br>
><br>
> -----------------------------------------------<br>
><br>
</div></div>> <log_view.txt><br>
<br>
</blockquote></div><br></div></div>