[Carpet] Synchronising, prolongating, and applying boundary conditions

[Erik Schnetter]

There exists a certain peculiarity with respect to boundary  
prolongation and applying boundary conditions in Carpet.  When using  
mesh refinement, it is necessary to apply boundary conditions after  
synchronising grid functions, not before.

The reason is the following.  There is unfortunately no central  
entity in Cactus which knows which grid points of a grid function are  
set by the boundary condition (which may be a symmetry condition).   
Carpet therefore does not know which points are boundary points, and  
it fills as many grid points as possible through interpolation when  
it applies boundary prolongation.  This will in general also set some  
fine grid boundary points.  Since interpolation is less accurate than  
applying the boundary condition directly, this will introduce some  
small error.  This error can be avoided by applying the boundary  
condition after synchronising instead of before.

When synchronising and applying boundary conditions in the "standard"  
way using the thorns Boundary and MoL, then the user usually  
"naturally" specifies the right thing.

Without mesh refinement, the order in which one synchronises and  
applies boundary conditions is in general irrelevant.

-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/20060802/20d6377a/attachment.pgp