[Carpet] Restriction and interpolation

[Erik Schnetter]

On Sep 1, 2006, at 15:37:44, Ian Hinder wrote:

> Hi,
>
> Is it true that quantities computed in the ANALYSIS bin are not
> restricted?  This seems to be the case both from looking at
> doc/internals and doc/scheduling, and also the output from my code.
> What is the rationale for this?

This is true.  The rationale is that the analysis bin is executed  
after restriction.  The time at which restriction occurs has a  
definite place in the Berger-Oliger time evolution algorithm --  
unlike boundary prolongation, which can happen at any time.  (This is  
partly only a design decision in Carpet.)

If you want to calculate norms or interpolate or look at reasonable I/ 
O of quantities, it is almost always necessary to calculate them in  
the evol bin, and to allocate three time levels for them.  Only then  
can you prolongate the mesh refinement boundaries, which requires  
interpolation in time.  If you want to restrict quantities, then  
there is no advantage in scheduling them in the analysis bin, since  
you very likely need to calculate them at all time steps anyway.

If you want to calculate a quantity only level-by-level, then  
restriction is not necessary (since the quantity is not evolved in  
time).  Using a more elaborate visualisation strategy is sufficient  
in this case.

> A slightly related question.  I assume that the interpolator uses the
> 'finest' points it can at a given spatial location.  When the
> interpolator asks Carpet for points, does Carpet ignore the contents
> of buffer zones, both when inner and outer buffer zones are being
> used?

It ignores outer buffer zones, but not inner buffer zones.

-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/20060903/64598498/attachment.pgp