[FLASH-USERS] Problems with set_parm_in_context

[Anshu Dubey]

Paul,

It has been a while since I have looked at FLASH2, so I may be wrong.
However, looking at the source code now, I see no evidence that
"mark_grid_refinement" is called from the default implementation of
init_from_checkpoint. It is called from "evolve", and when you use
"init_from_scratch", then there are calls to mesh_refine_derefine. I
do know more about what we do in FLASH 3, and since most of the logic
is similar, I am sure that FLASH 2 would be doing similar work at
initialization.

When you start from scratch, there is a process of refinement in a
loop to create the initial grid. There is no immediate need to refine
when you start from checkpoint. We do use update_refinement call, but
that is just to redistribute the blocks, there is no actual refinement
done. The code in FLASH 2 looks like it actually carries out the
distribution of blocks explicitly rather than handing control back to
paramesh.

Again, I could be wrong about all of this from being out of touch with
FLASH 2, but I do recommend that you make sure that
"mark_grid_refinement" is actually called from restart.

Best,
Anshu

On Thu, Jan 21, 2010 at 12:45 PM, Paul Goodall <ptg at astro.ox.ac.uk> wrote:
> Dear Flash-users,
>
> I don't think this qualifies as a bug, more likely just something I've missed out.
>
> I'm using FLASH2.5 (haven't yet had the the chance to upgrade) and having really weird problems with saving parameters to the database.
>
> In the mark_grid_refinement.F90 subroutine of my 'problem' setups directory, I calculate various quantities which are essential later on in the simulation, and save them to the database using:
>
> call set_parm_in_context(global_parm_context, 'parmname', parmval)
>
> (N.B I don't use 'add_parm_to_context' because the params in question are already initialised in Config)
>
> These values are calculated and saved during the IF(firstcall) loop.
> This is the bit were things become strange...
>
> If I do a new run (i.e. restart = .false. in flash.par) then everything works brilliantly and all the calculated quantities are saved to the database for later use.
>
> However, if I restart from a checkpoint, only *some* of the calculated quantities are saved to the database in the above way.  (i.e. when I call the parms later on in the source_terms/, some of them have the correct saved value, and some of them have reverted to the default value given in Config).  I know that the mark_grid_refinement.F90 is still called during a restart because I get it to print values to the screen/output files in both new starts and restarts such as:
>
> call get_parm_from_context(global_parm_context, 'parmname', parmval)
> print *,"Initial value: ", parmval
> (** some calculation involving parmval **)
> print *,"New value1: ", parmval
> call set_parm_in_context(global_parm_context, 'parmname', parmval)
> call get_parm_from_context(global_parm_context, 'parmname', parmval)
> print *,"New value2: ", parmval
>
> Where "New value1" and "New value2" being the same should surely indicate that the save to the database was successful? (I know that the above example is a bit dramatic, but its what I've resorted to to try and locate the problem).
>
> Two things about this are strange to me:
> 1) Why does this differ between a new start and a restart?
> 2) Why are some params saved to the database, and others not, when they each have the same type (e.g REAL).
>
> Any help with this (or if anyone has experience similar), would be really appreciated,
>
> cheers,
>
> Paul Goodall
> ========================
> Astrophysics D.Phil Student
> University of Oxford.
> ========================
> Denys Wilkinson Building
> Keble Road
> Oxford
> OX1 3RH
> +44 (0)1865-2-83011
> http://astroweb1.physics.ox.ac.uk/~Goodall/
> ========================
>
>