[aspect-devel] Purpose of denormalize_pressure?

[Timo Heister]

I think it might not be strictly necessary, but it might be safer to
do. Here are the places where it is used:
1. Computing the nonlinear residual (the double returned by
solve_stokes() before solving). It is only used multiplied by B^T, so
the constant shouldn't matter.
2. In computing the relative tolerance for the linear stokes solve.
Again we multiply by B^T.
3. The initial guess for the FGMRES solver itself. If you look at how
GMRES is implemented, the initial guess should only be used after
multiplying by the system matrix.

It is worth thinking about whether some of the operations we also
apply to the pressure (pressure scaling, applying/set_zero of hanging
node constraints) might interfere with the arguments above. Finally,
while the constant pressure is in the null space of our operator, it
could be that small numerical artifacts could influence this constant
(for example coming from an AMG preconditioner).

Again, I think we should be fine without it, and maybe we did one of
the three things above slightly differently and it mattered at some
point. Did you try removing the denormalization and run the tests?

On Tue, Sep 20, 2016 at 8:44 PM, Wolfgang Bangerth <bangerth at tamu.edu> wrote:
> On 09/20/2016 06:13 PM, Jonathan Perry-Houts wrote:
>>
>> According to the git "blame" feature, that function was introduced here:
>>
>> https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_geodynamics_aspect_commit_758005767378da1b08a2ebbce8ae95ac7daeef71&d=CwIGaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=R5lvg9JC99XvuTgScgbY_QFS80R7PEA2q0EPwDy7VQw&m=1Tsi1vCm0yh5oKTRnezG_zrJQHT4fH2FU_-SQpkDn2A&s=L338ZtSagzufKgzBfU7nHAqhkoONjJfQ-8zVwDUqWF8&e=
>> hope that helps!
>
>
> Thanks for digging this up, Jonathan.
>
> The patch is unfortunately not exactly clear how specifically it fixes the
> nonlinear solvers :-( But it confirms my suspicion that it has something to
> do with the nonlinear solvers.
>
> Best
>  W.
> --
> ------------------------------------------------------------------------
> Wolfgang Bangerth          email:                 bangerth at colostate.edu
>                            www:
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.math.colostate.edu_-7Ebangerth_&d=CwIGaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=R5lvg9JC99XvuTgScgbY_QFS80R7PEA2q0EPwDy7VQw&m=1Tsi1vCm0yh5oKTRnezG_zrJQHT4fH2FU_-SQpkDn2A&s=IW8NMkJOwHoFVlyqh6dAHZ-to40ow8sM3qk9iQnU1Gs&e=
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.geodynamics.org_cgi-2Dbin_mailman_listinfo_aspect-2Ddevel&d=CwIGaQ&c=Ngd-ta5yRYsqeUsEDgxhcqsYYY1Xs5ogLxWPA_2Wlc4&r=R5lvg9JC99XvuTgScgbY_QFS80R7PEA2q0EPwDy7VQw&m=1Tsi1vCm0yh5oKTRnezG_zrJQHT4fH2FU_-SQpkDn2A&s=I4O2i0ofolY8q2tFYyuT5NsmoPfoq2p1kb0TA3pEUhM&e=

-- 
Timo Heister
http://www.math.clemson.edu/~heister/