[Carpet] Scheduling pitfalls

[Erik Schnetter]

Christian Ott just discovered a scheduling error in thorn AEIThorns/ 
Dissipation.  I would like to take this opportunity to mention two  
rules of thumbs for scheduling routines for mesh refinement:

1. Routines which set up grid functions which are not evolved in time  
and which have only a single time level have to be scheduled also in  
the postregrid bin.  For example, routines that set up coordinate  
arrays or an excision mask fall into this class.  The rule of thumb  
is that routines that are scheduled in the basegrid bin should also  
be scheduled in the postregrid bin.

2. Routines which apply boundary conditions have to be scheduled also  
in both the postrestrict and the postregrid bin.  The reason is that  
boundaries cannot be filled by interpolation (since they are too  
close to the boundary) and therefore have to be set by the user.   
People may be tempted to say "but I'm not planning to have any mesh  
refinement near my boundaries!"  However, remember that people often  
want to have mesh refinement near symmetry boundaries.  The rule of  
thumb is that routines that are scheduled in the group MoL_PostStep  
should also be scheduled in the postrestrict and postregrid bins.

If you don't do this, then you will end up with grid functions that  
either have wrong boundaries or are left uninitialised.  This is  
often difficult to detect.

-erik

PS: Don't get confused by the many post... bins and groups: There are  
postinitial, poststep, MoL_PostStep, postrestrict, postregrid.

-- 
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/20060801/4fbd02d7/attachment.pgp