[Carpet] Regridding and scheduling in new version of Carpet

[Erik Schnetter]

On Mar 3, 2008, at 13:41:39, Ian Hinder wrote:

> Hi,
>
> I have a question about scheduling in the new version of Carpet.  I  
> have
> a thorn which sets up the CarpetRegrid coordinates string based on a
> spherical surface.
>
> In carpet-stable-3, I set the spherical surface in BASEGRID, and set  
> the
> coordinates string in PREREGRID.  This appears to work.
>
> In the development version of Carpet, things have changed.  There is  
> now
> a schedule bin PREREGRIDINITIAL, and regridding now seems to happen
> *before* BASEGRID.  The output from Cactus says:
>
>  Initialisation
>    [CCTK_PREREGRIDINITIAL]
>    Set up grid hierarchy
>    [CCTK_POSTREGRIDINITIAL]
>    [CCTK_BASEGRID]
>      [CCTK_INITIAL]
>      Restrict from finer grids
>      [CCTK_POSTRESTRICTINITIAL]
>      [CCTK_POSTINITIAL]
>
> So it seems that I have to set the coordinates string in
> PREREGRIDINITIAL in order to have the grid hierarchy set up correctly
> when the initial regridding happens.

Yes, this is correct.

> However, the information required
> to do this comes from a spherical surface.  I could set this spherical
> surface in PREREGRIDINITIAL as well, but then it will get  
> overwritten by
> SphericalSurface_Setup which is called in BASEGRID.  Have I understood
> this correctly?

Spherical surfaces are not supposed to contain useful information  
before POSTSTEP has been run.  This is because they may want to  
contain information about the state of the simulation, which is in  
general unknown before POSTSTEP.

It may be possible to initialise spherical surfaces before basegrid,  
but that would complicate things significantly, since they would  
potentially change later when apparent horizons are found.

The easiest solution would be to set the grid hierarchy in different  
ways in PREREGRIDINITIAL and in PREREGRID.  I find that setting up the  
initial grid hierarchy and modifying the grid hierarchy later on are  
sufficiently different that using the same method doesn't quite work.

> The documentation in Carpet/doc/scheduling.tex says that the bins are
> run in the order
>
>    [CCTK_BASEGRID]
>      [CCTK_INITIAL]
>    [CCTK_PREREGRIDINITIAL]
> 	Regrid
>    [CCTK_POSTREGRIDINITIAL]
>
> and this seems to me to make much more sense, but I understand if it  
> is
> out of date.

Yes, this is out of date.  Could you correct it?

> As an aside, I note that CarpetTracker, which tells CarpetRegrid2  
> where
> to put its grids based on a spherical surface, is scheduled only in
> PREREGRID.  I don't see how this can work?


This works only during time evolution.  For the initial grid hierarchy  
it is not possible to locate horizons (because the grid hierarchy is  
not there yet), hence the initial location is the one specified in the  
parameter file.

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : /archives/developers/attachments/20080303/a25d9fd7/attachment.pgp