[petsc-users] -log_view hangs unexpectedly // how to optimize my kspsolve

[Matthew Knepley]

On Sat, Jan 7, 2017 at 6:47 PM, Manuel Valera <mvalera at mail.sdsu.edu> wrote:

> Awesome Matt and Jed,
>
> The GCR is used because the matrix is not invertible and because this was
> the algorithm that the previous library used,
>
> The Preconditioned im aiming to use is multigrid, i thought i configured
> the hypre-boomerAmg solver for this, but i agree in that it doesn't show in
> the log anywhere, how can i be sure is being used ? i sent -ksp_view log
> before in this thread
>

Yes, that says you are using Block Jacobi/ILU. Try running with

 -pc_type gamg

which is algebraic multigrid. That will give good iteration counts, but not
be as efficient as geometric MG. Once
that works, you can switch over to geometric.


> I had a problem with the matrix block sizes so i couldn't make the petsc
> native multigrid solver to work,
>

Please send the error.


> This is a nonhidrostatic pressure solver, it is an elliptic problem so
> multigrid is a must,
>

Yes, MG is the way to go.

   Matt


> Regards,
>
> Manuel
>
> On Sat, Jan 7, 2017 at 4:34 PM, Jed Brown <jed at jedbrown.org> wrote:
>
>> Manuel Valera <mvalera at mail.sdsu.edu> writes:
>>
>> > I was able to find the bug, it was the outer loop bound in the same
>> fashion
>> > than before, my -log_view is this :
>> [...]
>> >       ##########################################################
>> >       #                                                        #
>> >       #                          WARNING!!!                    #
>> >       #                                                        #
>> >       #   This code was compiled with a debugging option,      #
>> >       #   To get timing results run ./configure                #
>> >       #   using --with-debugging=no, the performance will      #
>> >       #   be generally two or three times faster.              #
>> >       #                                                        #
>> >       ##########################################################
>>
>> The above isn't a joke.
>>
>> > VecMDot              525 1.0 1.7089e+00 1.7 1.48e+09 1.0 0.0e+00 0.0e+00
>> > 1.0e+03  7 17  0  0  6   7 17  0  0  6  1735
>> >
>> > VecMAXPY            1050 1.0 2.3646e+00 1.1 2.97e+09 1.0 0.0e+00 0.0e+00
>> > 0.0e+00 11 34  0  0  0  11 34  0  0  0  2508
>>
>> You are spending about a third of the solve time doing vector work.
>> What is your reason for using GCR?
>>
>> > KSPSolve               5 1.0 1.2218e+01 1.0 8.66e+09 1.0 1.1e+03 2.0e+04
>> > 1.9e+04 59100 99 43 99  59100 99 43 99  1418
>> >
>> > PCSetUp                3 1.0 1.7993e+00 1.0 1.27e+07 1.0 0.0e+00 0.0e+00
>> > 1.0e+01  8  0  0  0  0   8  0  0  0  0    14
>> >
>> > PCSetUpOnBlocks        5 1.0 1.9013e-01 1.7 1.27e+07 1.0 0.0e+00 0.0e+00
>> > 0.0e+00  1  0  0  0  0   1  0  0  0  0   134
>> >
>> > PCApply              546 1.0 3.8320e+00 1.1 1.77e+09 1.0 0.0e+00 0.0e+00
>> > 1.0e+00 18 20  0  0  0  18 20  0  0  0   925
>>
>> To make a big improvement, you'll need a better preconditioner.  What
>> kind of problem is this?
>>
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20170107/5e5c4612/attachment.html>