[Carpet] Postregrid

[Erik Schnetter]

On Mar 17, 2008, at 16:47:10, Ian Hinder wrote:

> Hi,
>
> In the notes for the new version of Carpet,
>
>  http://www.carpetcode.org/version-4.html
>
> it says:
>
> "There are new schedule bins CCTK_PREREGRIDINITIAL  and
> CCTK_POSTREGRIDINITIAL, called before and after regridding during
> initial data setup, respectively. As a rule of thumb, routines  
> scheduled
> in CCTK_BASEGRID must also be scheduled in both CCTK_POSTREGRID  and
> CCTK_POSTREGRIDINITIAL. You can use poisoning (see above) to catch
> errors in your schedule.ccl files."
>
> I don't understand this.  What is the purpose of the BASEGRID bin?
> In the Cactus documentation, it says:
>
>        This timebin is executed very early after a driver thorn
>        constructs grid; this bin should only be used to set up
>        coordinate systems on the newly created grids.
>
> Concerning POSTREGRID, it says:
>
>        This timebin is used in mesh refinement settings.  It is
>        ignored for unigrid runs.  This bin is executed whenever the
>        grid hierarchy or patch setup has changed during evolution;
>        see {\tt CCTK\_POSTREGRIDINITIAL}.  It is, e.g.,
>        necessary to re-apply the boundary conditions or recalculate
>        the grid points' coordinates after every changing the grid
>        hierarchy.
>
> So it seems that only the coordinates are supposed to be set in
> BASEGRID.  However, many other functions are scheduled in BASEGRID,  
> for
> example
>
>      ADMBase: Set the shift_state variable to 1
>      CarpetIOASCII: Initialisation routine
>      CarpetIOBasic: Initialisation routine
>      CarpetIOHDF5: Initialisation routine
>      CarpetIOScalar: Initialisation routine
>      CartGrid3D: Set up ranges for spatial 3D Cartesian coordinates
>      CartGrid3D: Set up spatial 3D Cartesian coordinates on the GH
>      SphericalSurface: Calculate surface coordinate descriptors
>      Dissipation: Ensure that there are enough ghost zones
>      GROUP MaskBase_SetupMask: Set up the weight function
>        CarpetReduce: Initialise the weight function
>        GROUP SetupMask: Set up the weight function (schedule other
> routines in here)
>          CarpetReduce: Set up the outer boundaries of the weight  
> function
>          CarpetReduce: Set up the weight function for the restriction
> regions
>      SymBase: Check whether the driver set up the grid consistently
>      Time: Initialise Time variables
>      Time: Set timestep based on Courant condition (courant_static)
>
>
>
> It seems that BASEGRID is a place where you can do initialization of
> grid variables which will happen in both initial data and recovery  
> runs,
> not just setting up the coordinates.  Is this correct?


This is correct.

My rule of thumb above is too broad.  Routines which are scheduled in  
basegrid and which initialise grid functions should very likely be  
scheduled in postregrid as well.  However, routines which initialise  
only grid scalars or grid arrays do not need to be scheduled in  
postregrid.

 From the list above, only the initialisation of the coordinates and  
the reduction weight should (and is) also be scheduled in postregrid.

-erik

-- 
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/20080317/a547624e/attachment-0001.pgp