<html><head></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><div>Also, note that PCMG can not create coarse grid spaces for an (MPI)AIJ matrix. If you use regular grids (DA?) then PETSc can construct geometric multigrid coarse grid spaces, although I don't know if PCMG will construct these for you (I don't think it will and I can see from your output that PCMG just used one grid). 'ml', hypre' and 'gamg' (a native AMG solver) will do real AMG solvers for you. All three can work on a similar class of problems.</div><div><br></div><div>Also, you mention that you have a condition number of 1.e20. That is astronomical for such a small problem. How did you compute that number? Do you know where the ill-conditioning comes from? Is this an elliptic operator?</div><div><br></div><div>Mark</div><div><br></div><br><div><div>On May 1, 2012, at 6:31 PM, Matthew Knepley wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div class="gmail_extra">On Tue, May 1, 2012 at 6:27 PM, Karthik Duraisamy <span dir="ltr"><<a href="mailto:dkarthik@stanford.edu" target="_blank">dkarthik@stanford.edu</a>></span> wrote:<br><div class="gmail_quote">
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">The following was output for the very first iteration whereas what I had attached earlier was output every iteration. I am still a bit perplexed because PCMG drops the residual like a rock (after the first few iterations whereas with no PCMG, it is very slow)<br>
</blockquote><div><br></div><div>Because the smoother IS the solver you were using before. Just like I said last time, what you are doing is</div><div>wrapping up the same solver you used before, sticking it in another GMRES loop, and only looking at the</div>
<div>outer loop. This has nothing to do with MG.</div><div><br></div><div> Matt</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
KSP Object: 8 MPI processes<br>
type: gmres<br>
GMRES: restart=100, using Classical (unmodified) Gram-Schmidt Orthogonalization with no iterative refinement<br>
GMRES: happy breakdown tolerance 1e-30<br>
maximum iterations=1, initial guess is zero<br>
using preconditioner applied to right hand side for initial guess<br>
tolerances: relative=0.01, absolute=1e-08, divergence=1e+10<br>
left preconditioning<br>
using DEFAULT norm type for convergence test<br>
PC Object: 8 MPI processes<br>
type: mg<br>
MG: type is MULTIPLICATIVE, levels=1 cycles=v<br>
Cycles per PCApply=1<br>
Not using Galerkin computed coarse grid matrices<br>
Coarse grid solver -- level -------------------------------<br>
KSP Object: (mg_levels_0_) 8 MPI processes<br>
type not yet set<br>
maximum iterations=1, initial guess is zero<br>
tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
left preconditioning<br>
using DEFAULT norm type for convergence test<br>
PC Object: (mg_levels_0_) 8 MPI processes<br>
type not yet set<br>
linear system matrix = precond matrix:<br>
Matrix Object: 8 MPI processes<br>
type: mpiaij<br>
rows=75000, cols=75000<br>
total: nonzeros=4427800, allocated nonzeros=4427800<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node (on process 0) routines: found 3476 nodes, limit used is 5<br>
<br>
<br>
----- Original Message -----<br>
From: "Matthew Knepley" <<a href="mailto:knepley@gmail.com">knepley@gmail.com</a>><br>
To: "PETSc users list" <<a href="mailto:petsc-users@mcs.anl.gov">petsc-users@mcs.anl.gov</a>><br>
Sent: Tuesday, May 1, 2012 3:22:56 PM<br>
Subject: Re: [petsc-users] Multigrid<br>
<br>
<br>
On Tue, May 1, 2012 at 6:18 PM, Karthik Duraisamy < <a href="mailto:dkarthik@stanford.edu">dkarthik@stanford.edu</a> > wrote:<br>
<br>
<br>
<br>
Hello,<br>
<br>
Sorry (and thanks for the reply). I've attached the no multigrid case. I didn't include it because (at least to the untrained eye, everything looks the same).<br>
<br>
<br>
<br>
Did you send all the output from the MG case? There must be a PC around it. By default its GMRES, so there would be<br>
an extra GMRES loop compared to the case without MG.<br>
<br>
<br>
Matt<br>
<br>
<br>
Regards,<br>
Karthik<br>
<br>
KSP Object: 8 MPI processes<br>
type: gmres<br>
GMRES: restart=100, using Classical (unmodified) Gram-Schmidt Orthogonalization with no iterative refinement<br>
GMRES: happy breakdown tolerance 1e-30<br>
maximum iterations=1<br>
using preconditioner applied to right hand side for initial guess<br>
tolerances: relative=1e-05, absolute=1e-50, divergence=1e+10<br>
left preconditioning<br>
using nonzero initial guess<br>
using PRECONDITIONED norm type for convergence test<br>
PC Object: 8 MPI processes<br>
type: bjacobi<br>
block Jacobi: number of blocks = 8<br>
Local solve is same for all blocks, in the following KSP and PC objects:<br>
KSP Object: (sub_) 1 MPI processes<br>
type: preonly<br>
maximum iterations=10000, initial guess is zero<br>
tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
left preconditioning<br>
using NONE norm type for convergence test<br>
PC Object: (sub_) 1 MPI processes<br>
type: ilu<br>
ILU: out-of-place factorization<br>
0 levels of fill<br>
tolerance for zero pivot 1e-12<br>
using diagonal shift to prevent zero pivot<br>
matrix ordering: natural<br>
factor fill ratio given 1, needed 1<br>
Factored matrix follows:<br>
Matrix Object: 1 MPI processes<br>
type: seqaij<br>
rows=9015, cols=9015<br>
package used to perform factorization: petsc<br>
total: nonzeros=517777, allocated nonzeros=517777<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node routines: found 3476 nodes, limit used is 5<br>
linear system matrix = precond matrix:<br>
Matrix Object: 1 MPI processes<br>
type: seqaij<br>
rows=9015, cols=9015<br>
total: nonzeros=517777, allocated nonzeros=517777<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node routines: found 3476 nodes, limit used is 5<br>
linear system matrix = precond matrix:<br>
Matrix Object: 8 MPI processes<br>
type: mpiaij<br>
rows=75000, cols=75000<br>
total: nonzeros=4427800, allocated nonzeros=4427800<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node (on process 0) routines: found 3476 nodes, limit used is 5<br>
<br>
<br>
----- Original Message -----<br>
From: "Matthew Knepley" < <a href="mailto:knepley@gmail.com">knepley@gmail.com</a> ><br>
To: "PETSc users list" < <a href="mailto:petsc-users@mcs.anl.gov">petsc-users@mcs.anl.gov</a> ><br>
Sent: Tuesday, May 1, 2012 3:15:14 PM<br>
Subject: Re: [petsc-users] Multigrid<br>
<br>
<br>
On Tue, May 1, 2012 at 6:12 PM, Karthik Duraisamy < <a href="mailto:dkarthik@stanford.edu">dkarthik@stanford.edu</a> > wrote:<br>
<br>
<br>
<br>
Hello Barry,<br>
<br>
Thank you for your super quick response. I have attached the output of ksp_view and it is practically the same as that when I don't use PCMG. The part I don't understand is how PCMG able to function at the zero grid level and still produce a much better convergence than when using the default PC. Is there any additional smoothing or interpolation going on?<br>
<br>
<br>
<br>
You only included one output, so I have no way of knowing what you used before. However, this is running GMRES/ILU.<br>
<br>
<br>
Also, for Algebraic Multigrid, would you recommend BoomerAMG or ML ?<br>
<br>
<br>
<br>
They are different algorithms. Its not possible to say generally that one is better. Try them both.<br>
<br>
<br>
Matt<br>
<br>
<br>
Best regards,<br>
Karthik.<br>
<br>
type: mg<br>
MG: type is MULTIPLICATIVE, levels=1 cycles=v<br>
Cycles per PCApply=1<br>
Not using Galerkin computed coarse grid matrices<br>
Coarse grid solver -- level -------------------------------<br>
KSP Object: (mg_levels_0_) 8 MPI processes<br>
type: gmres<br>
GMRES: restart=30, using Classical (unmodified) Gram-Schmidt Orthogonalization with no iterative refinement<br>
GMRES: happy breakdown tolerance 1e-30<br>
maximum iterations=1, initial guess is zero<br>
tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
left preconditioning<br>
using PRECONDITIONED norm type for convergence test<br>
PC Object: (mg_levels_0_) 8 MPI processes<br>
type: bjacobi<br>
block Jacobi: number of blocks = 8<br>
Local solve is same for all blocks, in the following KSP and PC objects:<br>
KSP Object: (mg_levels_0_sub_) 1 MPI processes<br>
type: preonly<br>
maximum iterations=10000, initial guess is zero<br>
tolerances: relative=1e-05, absolute=1e-50, divergence=10000<br>
left preconditioning<br>
using NONE norm type for convergence test<br>
PC Object: (mg_levels_0_sub_) 1 MPI processes<br>
type: ilu<br>
ILU: out-of-place factorization<br>
0 levels of fill<br>
tolerance for zero pivot 1e-12<br>
using diagonal shift to prevent zero pivot<br>
matrix ordering: natural<br>
factor fill ratio given 1, needed 1<br>
Factored matrix follows:<br>
Matrix Object: 1 MPI processes<br>
type: seqaij<br>
rows=9015, cols=9015<br>
package used to perform factorization: petsc<br>
total: nonzeros=517777, allocated nonzeros=517777<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node routines: found 3476 nodes, limit used is 5<br>
linear system matrix = precond matrix:<br>
Matrix Object: 1 MPI processes<br>
type: seqaij<br>
rows=9015, cols=9015<br>
total: nonzeros=517777, allocated nonzeros=517777<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node routines: found 3476 nodes, limit used is 5<br>
linear system matrix = precond matrix:<br>
Matrix Object: 8 MPI processes<br>
type: mpiaij<br>
rows=75000, cols=75000<br>
total: nonzeros=4427800, allocated nonzeros=4427800<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node (on process 0) routines: found 3476 nodes, limit used is 5<br>
linear system matrix = precond matrix:<br>
Matrix Object: 8 MPI processes<br>
type: mpiaij<br>
rows=75000, cols=75000<br>
total: nonzeros=4427800, allocated nonzeros=4427800<br>
total number of mallocs used during MatSetValues calls =0<br>
using I-node (on process 0) routines: found 3476 nodes, limit used is 5<br>
<br>
<br>
<br>
----- Original Message -----<br>
From: "Barry Smith" < <a href="mailto:bsmith@mcs.anl.gov">bsmith@mcs.anl.gov</a> ><br>
To: "PETSc users list" < <a href="mailto:petsc-users@mcs.anl.gov">petsc-users@mcs.anl.gov</a> ><br>
Sent: Tuesday, May 1, 2012 1:39:26 PM<br>
Subject: Re: [petsc-users] Multigrid<br>
<br>
<br>
On May 1, 2012, at 3:37 PM, Karthik Duraisamy wrote:<br>
<br>
> Hello,<br>
><br>
> I have been using PETSc for a couple of years with good success, but lately as my linear problems have become stiffer (condition numbers of the order of 1.e20), I am looking to use better preconditioners. I tried using PCMG with all the default options (i.e., I just specified my preconditioner as PCMG and did not add any options to it) and I am immediately seeing better convergence.<br>
><br>
> What I am not sure of is why? I would like to know more about the default parameters (the manual is not very explicit) and more importantly, want to know why it is working even when I haven't specified any grid levels and coarse grid operators. Any<br>
> help in this regard will be appreciated.<br>
<br>
First run with -ksp_view to see what solver it is actually using.<br>
<br>
Barry<br>
<br>
><br>
> Also, ultimately I want to use algebraic multigrid so is PCML a better option than BoomerAMG? I tried BoomerAMG with mixed results.<br>
><br>
> Thanks,<br>
> Karthik<br>
><br>
><br>
><br>
> --<br>
><br>
> =======================================<br>
> Karthik Duraisamy<br>
> Assistant Professor (Consulting)<br>
> Durand Building Rm 357<br>
> Dept of Aeronautics and Astronautics<br>
> Stanford University<br>
> Stanford CA 94305<br>
><br>
> Phone: <a href="tel:650-721-2835" value="+16507212835">650-721-2835</a><br>
> Web: <a href="http://www.stanford.edu/~dkarthik" target="_blank">www.stanford.edu/~dkarthik</a><br>
> =======================================<br>
<span class="HOEnZb"><font color="#888888"><br>
<br>
<br>
<br>
<br>
--<br>
What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>
-- Norbert Wiener<br>
<br>
<br>
<br>
<br>
--<br>
What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>
-- Norbert Wiener<br>
</font></span></blockquote></div><br><br clear="all"><div><br></div>-- <br>What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.<br>
-- Norbert Wiener<br>
</div>
</blockquote></div><br></body></html>