[FLASH-USERS] Typo in unsplit_rad Hydro_computeDT

[Mark Richardson]

Hi Klaus,

  That's exactly what was happening! I've implemented your ChildOrder
suggestion (switching HYPRE and Multigrid ranking), and now the potential
is correct, and I am not having the non-convergence issues anymore.

  Also, thank you for the comments on the boundary conditions. I've always
been a bit uncertain how these can be paired together, and your email
cleared that up.

Take care,
  -Mark


On 3 March 2016 at 19:45, Klaus Weide <klaus at flash.uchicago.edu> wrote:

> Hello Mark,
>
> On going back to one of your older messages, I noticed something
> suspicious in your log file excerpt:
>
>
> On Mon, 15 Feb 2016, Mark Richardson wrote:
> > I'm having one other problem, which is maybe related. Originally I was
> > running this with a self-made Eos (Tillotson for planetary bodies, that
> > includes radiative transfer). I was getting the following errors
> > immediately after initialisation:
> >
> >  [ 02-15-2016  09:57:48.810 ] [gr_hypreInit]: HYPRE_RELEASE_NAME=hypre,
> > HYPRE_RELEASE_VERSION=2.10.0b, HYPRE_RELEASE_DATE=2015/01/22
> >
> >  [ 02-15-2016  09:57:48.812 ] memory: /proc vsize    (MB):     2006.58
> > (min)       2008.35 (max)       2007.17 (avg)
> >  [ 02-15-2016  09:57:48.813 ] memory: /proc rss      (MB):     1284.28
> > (min)       1286.47 (max)       1285.21 (avg)
> >  [ 02-15-2016  09:57:48.815 ] memory: /proc vsize    (MB):     2006.58
> > (min)       2008.35 (max)       2007.17 (avg)
> >  [ 02-15-2016  09:57:48.816 ] memory: /proc rss      (MB):     1284.37
> > (min)       1286.56 (max)       1285.29 (avg)
> >
> >  [ 02-15-2016  09:57:58.820 ] [gr_hypreSolve]: Nonconv./failure: ierr=
> >       0
> >  [ 02-15-2016  09:57:58.820 ] [gr_hypreSolve]: Nonconv./failure:
> component=
> >           0
> >  [ 02-15-2016  09:57:58.820 ] [gr_hypreSolve]: Nonconv./failure:
> converged=
> > F
> >  [ 02-15-2016  09:57:58.820 ] [gr_hypreSolve]: Nonconv./failure: |initial
> > RHS|^2=  1.2520E+35
> >  [ 02-15-2016  09:57:58.820 ] [gr_hypreSolve]: Nonconv./failure:
> > num_iterations=         500
> >  [ 02-15-2016  09:57:58.820 ] [gr_hypreSolve]: Nonconv./failure:
> > final_res_norm=   88143026.2676523
>
> The fact that these messages appear this early - before writing of the
> initial checkpoint file, before the message "Entering evolution loop" -
> indicates that they come from some call from Driver_initFlash, not
> Driver_evolveFlash. Unless you have made significant changes, there should
> be no calls to Diffuse_solveScalar from Driver_initFlash (directly or
> indirectly). But there IS an initial call to
> Gravity_potentialListOfBlocks. So I strongly suspect that your setup is
> using GridSolvers/HYPRE/Grid_solvePoisson.F90 as the Poisson solver
> implementation, rather than the intended (I assume)
> GridSolvers/Multigrid/Grid_solvePoisson.F90.
>
> Q1. Am I correct to assume that you are intending to use Multigrid as
> Poisson solver for Gravity?
>
> Q2. Can you confirm that Grid_solvePoisson.F90 in your object directory
> was pointing to GridSolvers/HYPRE/Grid_solvePoisson.F90, not
> GridSolvers/Multigrid/Grid_solvePoisson.F90 as it should?
>
> Q3. Is the situation still the same in your most recent configuration?
>
> If the answers are Y/Y/Y, the solution may be simple - try the following:
> in source/Grid/GridSolvers/Config, change the line that begins with
>
>  CHILDORDER
>
> to
>
>  CHILDORDER Pfft HYPRE Multigrid
>
> The important part here is changing the relative order of Multigrid and
> HYPRE, so Multigrid comes later.
> (I have omitted various alternative implementations from the directive
> that are not included in the released code anyway.)
>
>
> Klaus
>



-- 

Mark Richardson
Beecroft Postdoctoral Fellow, Department of Physics
Denys Wilkinson Building - 555F
University of Oxford
Mark.Richardson at physics.ox.ac.uk
+44 1865 280793
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160305/4d9f0839/attachment.htm>