[Carpet] [Cluster-users] job aborts with wierd error msg when Carpet does 1st recompose

[Erik Schnetter]

On Sep 27, 2006, at 10:22:34, Jonathan Thornburg wrote:

> Hi, Erik,
>
> On Wed, 27 Sep 2006, Erik Schnetter wrote:
>> This error is caused by
>>
>> Carpet::prolongation_order_space = 4
>>
>> You are selecting 4th order prolongation operators.  These do not  
>> exist.  You
>> need to choose either 3rd order or 5th order operators.
>
> Ok.
>
> For a system of PDEs (BSSN) which is 1st order in time and 2nd order
> in space, do I understand correctly that I need 3rd order prolongation
> to get overall 2nd order accuracy (assuming 2nd order FD in physics  
> thorns),
> and 5th order to get overall 4th order accuracy (assuming 4th order  
> FD in
> physics thorns)?

In principle, using order p-1 on the boundary can give you order p  
globally.  In practice this depends on the ratio of boundary points  
to interior points, and that ratio can -- in practice -- be  
surprisingly bad.  As far as I know, this is "folklore" knowledge,  
and I don't know the details under which this is true.  Since  
refinement boundaries have a large constant in front of that error  
term anyway, people usually use order p+1 on the boundary.

-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/20060927/f0684c38/attachment.pgp