[Carpet] regridding?

Erik Schnetter schnetter at cct.lsu.edu
Tue May 9 16:44:37 CEST 2006


On May 5, 2006, at 09:25:57, Bela Szilagyi wrote:

> Cristian Ott and I have found a feature of Carpet that could be  
> regarded as a
> bug.
>
> Say you are on refinement level 3 out of 3.  You ask for a regrid  
> by setting
> CarpetRegrid::coordinates, and your new parameter value involves a  
> change of
> refinement level 2.
>
> Carpet will regrid and schedule PostRegrid for refinement level 3  
> but not for
> the coarser levels.
>
> I would propose one of the following:
>
> 1. explicitely forbid requests of refinement level changes of a  
> coarser grid
> than the current one
>
> 2. delay these requests till evolution gets to the appropriate  
> coarser grid
>
> 3. schedule postregrid for the coarser levels as well
>
> 4. schedule postregrid for all levels at any regrid event.
>
> 5. whatever...

We are doing "whatever" already!  I'm glad we could be of service.

Seriously.  Carpet itself allows you to change the complete grid  
hierarchy at any time, simply because there is no technical reason to  
disallow that.  Whether it makes sense to do so, or whether you can  
get the desired kind of convergence depends on your algorithm.

I think therefore that (3) makes the most sense.  Which is the same  
as (4), so you'll get twice your money worth.  The postregrid bin  
should only be doing things that are idempotent, such as apply  
boundary conditions, so that running this bin too often should not  
matter.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>

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: 186 bytes
Desc: This is a digitally signed message part
Url : /archives/developers/attachments/20060509/ae55679b/attachment.pgp 


More information about the developers mailing list